Originally published on Towards AI the World’s Leading AI and Technology News and Media Company. If you are building an AI -related product or service, we invite you to consider becoming an AI sponsor . At Towards AI, we help scale AI and technology startups. Let us help you unleash your technology to the masses. This blog post is all about what we look for while making friends! Jokes apart, in this project, we will learn how to compute the similarity between a search query and a database of texts. We will also rank all the data according to their similarity score and retrieve the most similar text with its index. Background
Before we jump into how the project works, it is of utmost importance to understand its applications and uses of the same. Practically, all day we run into similarity algorithms , some advanced and some very basic and useful, such as the cosine similarity. Techno-legal domain data, such as patents, work a lot with this task statement to conduct searches for overlapping technologies or inventions. While search methods can either be corpus-based, the ones that rely on the larger corpora of information to draw semantics between the concepts, or, deep learning based, where neural networks are used to compute the embeddings as well as the distance. [1]
In this blog, we will discuss one example from both these methods as we understand how cosine similarity helps us in leveraging the distance information between these two different embedding approaches. Dataset & Preprocessing
In our experiments, we look into the Stack Overflow Questions Dataset where we attempt to find similar questions to our query. The dataset contains 60,000 Stack Overflow questions from 2016–2020. We use the function read_sc to read our search criteria which can be either entered as a string or as a path to the text file containing it, and the function is clean to load the input file into a dataframe . Preprocessing the text plays a key role for corpus-based embeddings such as Word2Vec and fastText. While domain-specific processing techniques such as introducing contraction of text or expansion of abbreviations improve the algorithms , our data is comparatively clean, and thus we opt for the standard NLP preprocessing techniques. The function below takes a string as input, removes punctuations and stopwords, and returns a list of tokens. Exploring The Embeddings
Sentence-BERT
The algorithm uses the paraphrase-MiniLM-L6-v2 by HuggingFace for generating the embeddings, which is used for tasks like clustering or semantic search since it maps phrases and paragraphs to a 384-dimensional dense vector space. Instead of using sentence transformers which add to the slug size while deploying the model, we pass our input to the transformer model and then apply the right pooling operation on top of the contextualized word embeddings. This transformer model was also versioned with DVC. FastText
But why use fastText when we already have all superior BERT-based embeddings in place, you ask? When it comes to long-form documents, nothing beats the good old fastText! The sentence-Bert embeddings have a common limitation value of 512-word pieces, and this length can not be increased. But when we use corpus-based embeddings such as Word2Vec or FastText, we do not have any such limitation. FastText operates at a finer level using character n-grams, where words are represented by the sum of the character n-gram vectors, in contrast to Word2Vec, which internally utilizes words to predict words. And thus, we never run into the classic ‘out of vocabulary’ error with FastText, and it works well for several languages for which the words are not present in its vocabulary. Calculating The Similarity With Cosine Similarity
Understanding Cosine Similarity ( Source )
As depicted in the illustration above, in the case of cosine similarity, we measure the angle between two vectors (or embeddings). There are numerous other similarity measures, such as the Euclidean distance and Jaccard similarity. However, cosine distance performs better than these as it is not just a measure of common words or magnitude of overlap of the concepts but considers the orientation of vectors in the embedding space and thus even if the documents are of incomparable sizes, it can accurately calculate their similarity. Above is the basic function to compute the cosine similarity between two vectors a and b for your reference. We explore Pytorch and Scipy-based cosine similarity functions in our code. We compute the embeddings of the search text as well as the sentences in our dataset , and for sentence-bert, we utilize the util.pytorch_cos_sim function and for fastText, we use scipy.spatial.distance.cosine to calculate similarity between them. Structuring the Code
We have a main.py file that takes the arguments such as model name, search criteria, a file containing all the data, name of the column from which you want to search the similar text. The main function calls one of the two classes defined for the BERT-based embedding and fastText embedding according to the argument you choose. Analyzing The Algorithms
As we are not dealing with the traditional supervised learning problem where we train our model to log the accuracy and hyperparameters, a challenge was to decide how to log and evaluate the performance. While the models return a ranked output file in the order of their similarity to the search text, I also decided to log the most similar sentence scored by both models individually and its index to understand if the models capture the semantics in a similar fashion. An interesting comparison parameter could be computation time as well. Let’s take a phrase and check the results! Phrase: Why is Java Optionals immutable? Both the algorithms return the same value where the selected match has nearly the same scores. Logs for Sent-BERT and FastText
Now the Streamlit App! pip install -r requirements.txt
Streamlit App for Similarity Calculation
Simply enter the string of text which you want to look for and the ‘.csv’ file in which you want to look into. You will need to enter the name of the column and make a choice between the two models! Here’s how the output looks :
Similarity results : Streamlit WebApp
Streamlit is easy to use, for example, to input the data and run the calculation function, we use :
Deploying The WebApp On AWS EC2 Instance
We decide to deploy our app in an AWS EC2 instance, and this blog is the holy grail for it. There are 5 steps to launch your EC2 instance :
Select an AMI (Amazon Machine Image), we go for the free tier eligible one! 2. Next, you need to choose an instance type. 3. Create and download an RSA type keypair. 4. Now, we need to modify Network Settings and create a new security group. You need to name your security group and then add two new rules with the ‘Custom TCP’ type and set the port range as 8501 for one rule and 8502 for another. 5. Now select the storage you require and launch the instance! Now your instance is running! I follow the easiest way to connect to this instance, and you can follow along. ????
Simply select the instance above and click to connect, and there you have the terminal right in front of you! If you have followed so far, you already have your model and data pushed to DVC (and the dagshub repository), and all we need to do now is to pull the model using a simple script that prepares the EC2 instance and launches our app. The script I followed is the same as the blog referred to above :

Searching For Semantic Similarity!

I track enterprise software application development & data management. Jun 14, 2022,
Snowflake
Developers develop software. But wider than this core truth, developers develop software in order to create code that is workable, executable, secure, functional, capable of being integrated and – ultimately – deployable and monetizable. They may also create it to fulfil some higher purpose for humanity and planet with an eye on sustainability, diversity and inclusivity… but essentially they want software to work and find its place with users. Snowflake has worked to try and reflect some (if not all) of these core realities with its latest product release. Snowflake’s Native Application Framework enables developer workflows and tasks by offering a platform for building, monetizing and deploying data-intensive applications in what the company calls the ‘data cloud’ space – a term it capitalizes to define its raison d’être and core technology proposition. With the new Native Application Framework , currently in private preview (meaning that a select group of pre-launch testers are finalizing its functionality nuances as part of a User Acceptance Testing (UAT) process), software developers can build applications and monetize them on Snowflake Marketplace , the firm’s own application portal. What are data-intensive applications? Users can securely install and run those applications, directly in their Snowflake cloud instances, reducing the need for data to be moved. This story gravitates around what Snowflake (and, for that matter, the lion’s share of other cloud-centric software specialists) call data-intensive applications. With every application on that planet, ever, only really existing as a result of the data that comprises it and that it holds, ingests, processes, augments, exchanges and delivers, the notion of a data-intensive application is meant to explain an application with an order of magnitude greater data workload, a potential conduit to big data analytics engines and an inherent architectural appreciation for massive data flows, in whatever form. Defining this higher data-intensive tier further, these are applications that work beyond megabytes and gigabytes into the realm of terabytes and petabytes (exabytes, zettabytes & yottabytes come next) with data that may exist in multiple formats at various stages of unstructured raw to structured states and through complex, tiered, multi-staged analytics pipelines – sometimes executing multiple processes in parallel and sometimes working on data jobs in more sequential steps. "Building on the idea that a data-intensive app isn't defined in terms of the volume of the data it utilizes, but the importance of the data to its core value, we view data-intensive apps as a newer class of apps that put data [combined with] Machine Learning (ML) at the heart of their value," clarified Snowflake, in an official press statement. MORE FOR YOU
This is what Snowflake is good at i.e. apps with a really big plate of data. The company insists that it is committed to providing its customers and partners with the best platform for building data-intensive applications and further states that Snowflake’s Native Application Framework is an important milestone in that journey. How does it all work? For software engineers shouldering the oft-immense tasks associated with building applications across the new cloud-connected landscape, this stuff will make sense – but we can break this technology down and explain it in comparatively digestible terms too. Software developers will use the Snowflake Native Application Framework to build applications using Snowflake functionalities such as stored procedures (software code scripts that can access, build, modify, delete parts of a database) and the related technology actions driven by user-defined functions (UDFs) and user-defined table functions (UDTFs) – all of which are packaged and componentized, approved and checked, tested and secured elements of data-centric application architecture that can be used and reused in use cases that repeatedly surface in many different scenarios. In the world of cloud, an application is not necessarily just an application; it needs to be an app that has all the requisite build mechanics available to it (think of it like the difference between driving and flying, you still move forwards, but there’s a whole different set of turbines powering your journey) – plus it needs to be able to be distributed (remember how we all work remotely around the planet now?) and deployable natively to the new world of cloud. This is what Snowflake has built i.e. a data-first software platform to make the new breed of data-intensive apps a reality. “Applications have been a part of Snowflake’s DNA from day one,” said Chris Child, senior director, product management, Snowflake. “The Snowflake Native Application Framework enables every customer to step into the future with a seamless way to build, distribute and use applications natively in [what we call] the Data Cloud.”
Capabilities here include the Streamlit integration for developing interactive customer interfaces and telemetry features including events and alerts for monitoring and troubleshooting are also under development. Streamlit is a software framework built to simplify and accelerate the creation of data applications and make creating web apps using the Python programming language faster and easier. Using this technology, data scientists are able to go from data and models to deployed apps (in a matter of hours) and only a handful of lines of code. Snowflake also points out the fact that its Native Application Framework accelerates application development by building on the organization's own high availability and disaster recovery, global collaboration capabilities and security posture so developers in the ‘data cloud’ can focus on functionality... and not operational burden. A new era of collaboration
By expanding Snowflake Marketplace to include applications that run natively in the data cloud, customers can get value faster by discovering and installing applications with just a few clicks. They can improve security and governance by using applications directly in their Snowflake instances. “Applications are an integral part of Snowflake’s expansive partner ecosystem. As of April 30, 2022, over 425 Powered by Snowflake partners have built applications that utilize Snowflake as their data infrastructure including Aladdin by BlackRock, Okta Inc. and Warner Music Group. With Snowflake, partners can extend their customer base and get the support they need to build better applications at scale,” stated the company, in a press statement. Together with Streamlit, acquired in March 2022 , Snowflake will further provide developers and data scientists with a single, powerful platform to discover, collaborate and build next generation applications. Developers can now build applications using the tools they love, with simplified data access and governance. Central themes & takeaways
In terms of central themes and takeaways here, we’re seeing Snowflake execute an arguably sophisticated (and unarguably technically complex) play that runs in line with many central themes surfacing in the enterprise software space today. Organizations want their developers to build applications with great functionality and an ability to scale, integrate, expand and evolve… all the time. So, logically, they need developers to be able to focus on front-end functions and smart services. In order to do that, they need to abstract, elevate and automate the software developer’s need to think about back-end engineering, data management and systems connectivity. We’re now going native in cloud i.e. cloud from first base ground zero – and Snowflake has quite directly named its new launch the Native Application Framework. The cloud could be coming out of the freezer, Snowflake may be hot.

Snowflake Defrosts The Data-Intensive Application

With its new Native Application Framework, Snowflake offers developers the ability to build and run applications on a common cloud-based data platform and leverage its Data Marketplace to monetize those apps. Thinkstock
Cloud-based data warehouse company Snowflake is switching gears. Launched in 2014 with a focus on disrupting the traditional data warehouse market and big-data analytics, the company is turning its attention toward application development. At its annual Snowflake Summit Tuesday, the company announced its Native Application Framework, which offers the ability to build and run applications inside its Data Cloud platform. The application framework, currently in private preview, allows developers to build applications and monetize them on Snowflake's Data Marketplace, said Chris Child, senior director of product development at Snowflake. Users of these applications can install and run them directly inside their Snowflake instances, he added. "We're sort of evolving from a data warehouse to a data platform to a data cloud. And now we want to evolve that from a data cloud to really an application cloud," said Chris Child, senior director of product development at Snowflake. About four years ago, with growing demand for analytics and the evolution of analytics-related architecture—such as separation of compute and storage—the company doubled its focus on making data collaboration easier within companies. That led Snowflake to recast its data warehouse, which evolved into its Data Cloud platform in 2019. Continued attempts to ease data sharing and collaboration resulted in the company's Data Marketplace. The marketplace enables data sharing within organizations, and also allows them to give other companies access to that data, charging them for usage. In March this year, the company launched two industry-specific data clouds, for the healthcare and retail sectors, with the intention of easing the way companies in those areas can share and monetize data sets. Build, run and monetize apps within Snowflake
Now, the rationale behind launching the Native Application Framework is reducing the time and effort needed to move data when building and running applications, said Tony Baer, principal analyst at dbInsights. This also means that developers will be able to build and run apps under the common Snowflake runtime, security, and governance umbrella, Baer added. In addition to offering Snowflake functionality such as stored procedures, user-defined functions (UDFs) and user-defined table functions (UDTFs) for developers, Snowflake will add integration with open source app framework Streamlit. The integration is meant to help develop interactive customer interfaces, as well as telemetry features such as events monitoring and alerts for troubleshooting. These features, according to Child, are still under development. Python -based Streamlit—aimed at machine learning and data scienc e engineering teams, to help visualize, mutate and share data—was acquired by Snowflake in March. Native Application Framework already has users
Several companies, including Capital One, LiveRamp and Informatica, have already used the Native Application Framework to build applications, Snowflake said. While Capital One's solution, called Slingshot, offers data management, LiveRamp and Informatica have developed applications for cloud cost management, identity resolution and data integration. The Snowflake Google Analytics and ServiceNow Connectors, both of which are in private preview, were also built using the Native Application Framework, Child said, adding that the Service Now Connector helps accelerate service desk ticket response. The revenue-sharing agreement for applications shared through the marketplace stands at 10% of the transaction value. "As dollar numbers get larger, then we'll have to have sort of like a grading scale. So, currently it starts at 10%, which is what we started with data sharing as well. And we will evolve it based on the use cases over time," said Christian Kleinerman, senior vice president of product at Snowflake. Unistore aims to unite analytical, transactional workloads
As organizations look to cut down time needed for application development, there is a need to solve the challenge of siloed transactional and analytical data, which in turn creates complexities when data has to be moved between systems. Targeting this specific issue, Snowflake on Tuesday also introduced a new product that brings analytical and transactional workloads together. Called Unistore, and currently in private preview, the system is designed to allow development teams to expand the Data Cloud to include transactional use cases such as application-state and data serving, the company said. In order to help Unistore build transactional applications, Snowflake is introducing a new capability called Hybrid Tables, which is also in private preview. Hybrid Tables offer fast single-row operations and allows customers to build transactional business applications directly on Snowflake, the company said, adding that the new tool will also allow customers to perform analytics on transactional data for immediate context, and join Hybrid Tables with existing Snowflake Tables for a holistic view across all data. Adobe is one of the early adopters of Unistore, Kleinerman said, adding that the company has been using the private preview of Hybrid Tables for its Adobe Campaigns application. Snowflake stresses transaction processing
The strategy behind the inclusion of more analytical workloads onto the Snowflake Data Cloud, according to Baer, is to solidify its position as the preferred destination platform for transaction processing use cases, embedding operational analytics, and not to compete head on with MySQL, Oracle, or SQL Server for pure OLTP (online transaction processing) applications. "From that standpoint, it is the mirror image of MongoDB, which is slowly adding analytics to its operational database. The goal is the same, but both players are coming at it from different starting points," Baer said. The other difference, according to Baer, is that Snowflake is further along in being able to embed operational analytics into transactions compared to MongoDB, where for now the process is still more primitive. The launch of the application framework along with Unistore will drive Snowflake's Data Marketplace strategy further, said Doug Henschen, principal analyst, Constellation Research, said. "This is a bold announcement whereby Snowflake is bidding for customers to bring more transactional data into applications in order run a broader set of applications on Snowflake while taking advantage of the Snowflake Data Marketplace to monetize those applications," Henschen said. Related:

Snowflake switches gears, from data warehouse to application cloud

How To Automate Ecommerce Category Page Creation With Python
0 0 8 minutes read
Clustering product inventory and automatically aligning SKUs to search demand is a great way to find opportunities to create new ecommerce categories. Niche category pages are a proven way for ecommerce sites to align with organic search demand while simultaneously assisting users in purchasing. If a site stocks a range of products and there is search demand, creating a dedicated landing page is an easy way to align with the demand. But how can SEO professionals find this opportunity? Sure, you can eyeball it, but you’ll usually leave a lot of opportunity on the table. This problem motivated me to script something in Python, which I’m sharing today in a simple to use Streamlit application. (No coding experience required!) The app linked above created the following output automatically using nothing more than two crawl exports! Screenshot from Microsoft Excel, May 2022
Notice how the suggested categories are automatically tied back to the existing parent category? Screenshot from Microsoft Excel, May 2022
The app even shows how many products are available to populate the category. Screenshot from Microsoft Excel, May 2022
Benefits And Uses
Increase the chance of relevant site links displaying underneath the parent category. Reduce CPCs to the landing page through increased relevancy. Potential to inform merchandising decisions. (If there is high search demand vs. low product count – there is a potential to widen the range.0
Mock-up Screenshot from Google Chrome, May 2022
Creating the suggested subcategories for the parent sofa category would align the site to an additional 3,500 searches per month with relatively little effort. Features
Match to a minimum of X products before recommending a category. Check similarity to an existing category (X % fuzzy match) before recommending a new category. Set minimum search volume/CPC cut-off for category suggestions. Supports search volume and CPC data from multiple countries. Getting Started/Prepping The Files
To use this app you need two things. At a high level, the goal is to crawl the target website with two custom extractions. The internal_html.csv report is exported, along with an inlinks.csv export. These exports are then uploaded to the Streamlit app, where the opportunities are processed. Crawl And Extraction Setup
When crawling the site, you’ll need to set two extractions in Screaming Frog – one to uniquely identify product pages and another to uniquely identify category pages. The Streamlit app understands the difference between the two types of pages when making recommendations for new pages. The trick is to find a unique element for each page type. (For a product page, this is usually the price or the returns policy, and for a category page, it’s usually a filter sort element.) Extracting The Unique Page Elements
Screaming Frog allows for custom extractions of content or code from a web page when crawled. This section may be daunting if you are unfamiliar with custom extractions, but it’s essential for getting the correct data into the Streamlit app. The goal is to end up with something looking like the below image. (A unique extraction for product and category pages with no overlap.) Screenshot from Screaming Frog SEO Spider, May 2022
The steps below walk you through manually extracting the price element for a product page. Then, repeat for a category page afterward. If you’re stuck or would like to read more about the web scraper tool in Screaming Frog, the official documentation is worth your time. Manually Extracting Page Elements
Let’s start by extracting a unique element only found on a product page (usually the price). Highlight the price element on the page with the mouse, right-click and choose Inspect. Screenshot from Google Chrome, May 2022
This will open up the elements window with the correct HTML line already selected. Right-click the pre-selected line and choose Copy > Copy selector. That’s it! Screenshot from Google Chrome, May 2022
Open Screaming Frog and paste the copied selector into the custom extraction section. (Configuration > Custom > Extraction). Screenshot from Screaming Frog SEO Spider, May 2022
Name the extractor as “product,” select the CSSPath drop down and choose Extract Text. Repeat the process to extract a unique element from a category page. It should look like this once completed for both product and category pages. Screenshot from Screaming Frog SEO Spider, May 2022
Finally, start the crawl. Screenshot from Screaming Frog SEO Spider, May 2022
Notice how the extractions are unique to each page type? Perfect. The script uses the extractor to identify the page type. Internally the app will convert the extractor to tags. (I mention this to stress that the extractors can be anything as long as they uniquely identify both page types.) Screenshot from Microsoft Excel, May 2022
Exporting The Files
Once the crawl has been completed, the last step is to export two types of CSV files. internal_html.csv. inlinks to product pages. Go to the Custom Extraction tab in Screaming Frog and highlight all URLs that have an extraction for products. (You will need to sort the column to group it.) Screenshot from Screaming Frog SEO Spider, May 2022
Lastly, right-click the product URLs, select Export, and then Inlinks. Screenshot from Screaming Frog SEO Spider, May 2022
You should now have a file called inlinks.csv. Finally, we just need to export the internal_html.csv file. Click the Internal tab, select HTML from the dropdown menu below and click on the adjacent Export button. Finally, choose the option to save the file as a .csv
Screenshot from Screaming Frog SEO Spider, May 2022
Congratulations! You are now ready to use the Streamlit app! Using The Streamlit App
Using the Streamlit app is relatively simple. The various options are set to reasonable defaults, but feel free to adjust the cut-offs to better suit your needs. I would highly recommend using a Keywords Everywhere API key (although it is not strictly necessary as this can be looked up manually later with an existing tool if preferred. (The script pre-qualifies opportunity by checking for search volume. If the key is missing, the final output will contain more irrelevant words.) If you want to use a key, this is the section on the left to pay attention to. Screenshot from Streamlit.io, May 2022
Once you have entered the API key and adjusted the cut-offs to your links, upload the inlinks.csv crawl. Screenshot from Streamlit.io, May 2022
Once complete, a new prompt will appear adjacent to it, prompting you to upload the internal_html.csv crawl file. Screenshot from Streamlit.io, May 2022
Finally, a new box will appear asking you to select the product and column names from the uploaded crawl file to be mapped correctly. Screenshot from Streamlit.io, May 2022
Click submit and the script will run. Once complete, you will see the following screen and can download a handy .csv export. Screenshot from Streamlit.io, May 2022
How The Script Works
Before we dive into the script’s output, it will help to explain what’s going on under the hood at a high level. At a glance:
Generate thousands of keywords by generating n-grams from product page H1 headings. Qualify keywords by checking whether the word is in an exact or fuzzy match in a product heading. Further qualify keywords by checking for search volume using the Keywords Everywhere API (optional but recommended). Check whether an existing category already exists using a fuzzy match (can find words out of order, different tenses, etc.). Uses the inlinks report to assign suggestions to a parent category automatically. N-gram Generation
The script creates hundreds of thousands of n-grams from the product page H1s, most of which are completely nonsensical. In my example for this article, n-grams generated 48,307 words – so this will need to be filtered! Screenshot from Microsoft Excel, May 2022
The first step in the filtering process is to check whether the keywords generated via n-grams are found at least X times within the product name column. (This can be in an exact or fuzzy match.) Anything not found is immediately discarded, which usually removes around 90% of the generated keywords. The second filtering stage is to check whether the remaining keywords have search demand. Any keywords without search demand are then discarded too. (This is why I recommend using the Keywords Everywhere API when running the script, which results in a more refined output.) It’s worth noting you can do this manually afterward by searching Semrush/Ahrefs etc., discarding any keywords without search volume, and running a VLOOKUP in Microsoft Excel. Cheaper if you have an existing subscription. Recommendations Tied To Specific Landing Pages
Once the keyword list has been filtered the script uses the inlinks report to tie the suggested subcategory back to the landing page. Earlier versions did not do this, but I realized that leveraging the inlinks.csv report meant it was possible. It really helps understand the context of the suggestion at a glance during QA. This is the reason the script requires two exports to work correctly. Limitations
Not checking search volumes will result in more results for QA. (Even if you don’t use the Keywords Everywhere API, I recommend shortlisting by filtering out 0 search volume afterward.) Some irrelevant keywords will have search volume and appear in the final report, even if keyword volume has been checked. Words will typically appear in the singular sense for the final output (because products are singular and categories are pluralized if they sell more than a single product). It’s easy enough to add an “s” to the end of the suggestion though. User Configurable Variables
But here is a run down if you’d like to tweak and experiment. Minimum products to match to (exact match) – The minimum number of products that must exist before suggesting the new category in an exact match. Minimum products to match to (fuzzy match) – The minimum number of products that must exist before suggesting the new category in a fuzzy match, (words can be found in any order). Minimum similarity to an existing category – This checks whether a category already exists in a fuzzy match before making the recommendation. The closer to 100 = stricter matching. Minimum CPC in $ – The minimum dollar amount of the suggested category keyword. (Requires the Keywords Everywhere API.) Minimum search volume – The minimum search volume of the suggested category keyword. (Requires Keywords Everywhere API.) Keywords Everywhere API key – Optional, but recommended. Used to pull in CPC/search volume data. (Useful for shortlisting categories.) Set the country to pull search data from – Country-specific search data is available. (Default is the USA.) Set the currency for CPC data – Country-specific CPC data is available. (Default USD.) Keep the longest word suggestion – With similar word suggestions, this option will keep the longest match. Enable fuzzy product matching – This will search for product names in a fuzzy match. (Words can be found out of order, recommended – but slow and CPU intensive.) Conclusion
With a small amount of preparation, it is possible to tap into a large amount of organic opportunity while improving the user experience. Although this script was created with an ecommerce focus, according to feedback, it works well for other site types such as job listing sites. So even if your site isn’t an ecommerce site, it’s still worth a try. Python enthusiast? More resources:

How To Automate Ecommerce Category Page Creation With Python

Build A Machine Learning Web App From Scratch
0 0 Less than a minute
Build a Machine Learning web application from scratch in Python with Streamlit. We use real world data to build a machine…
0 0 Less than a minute
Be the first to know the latest! Enter your Email address
Comment *
 − 
About Us
We successfully fused Technology and Audiovisual Entertainment to give you a very exciting experience on our website. Take advantage of our daily handpicked information and resources to widen your knowledge and build your skills. Most Viewed Posts

Build A Machine Learning Web App From Scratch

Zoox

Ditto

Google

Trouvé Media

Yale University

Carnegie Mellon University

Funding Date

Pre-Money Valuation

Post-Money Valuation

Amount Raised

Shares Authorized

Issuance Price

Dividend Rate

Liquidation Preferences

Liquidation Price

Participation

Conversion Price

Anti Dilution

General Voting

Board Voting

Par Value

Snowflake acquires Streamlit for $800M. Streamlit competitors include Plotly.

This Week In AI: Fintech Startup Pagaya Raises $102M, OpenAI Debuts Image Generation Model, Amazon Launches Social Distancing Monitoring Tool

This Week In AI: Tesla Acquires Computer Vision Startup, Microsoft Partners With Novartis, Apple Patents Digital Avatar Tool

This collection includes startups selling AI SaaS, using AI algorithms to develop their core products, and those developing hardware to support AI workloads.

Artificial Intelligence

Acquired

Seed VC

Series A

Series B

Streamlit is an app framework specifically for machine learning and data science teams. Streamlit aims to turn data scripts into shareable web apps in pure Python.On March 2nd, 2022, Streamlit was acquired by Snowflake.

Streamlit

Snowflake Computing

Name	Work History	Title	Status
Amanda Kelly	Zoox, Ditto, Google, and Trouvé Media	Founder, Chief Operating Officer	Current
Thiago Teixeira	Google, and Yale University	Founder	Current
Adrien Treuille	Zoox, Google, and Carnegie Mellon University	Founder, Chief Executive Officer	Current
Mark Overington		President, Senior Vice President	Current

Name	Amanda Kelly	Thiago Teixeira	Adrien Treuille	Mark Overington
Work History	Zoox, Ditto, Google, and Trouvé Media	Google, and Yale University	Zoox, Google, and Carnegie Mellon University
Title	Founder, Chief Operating Officer	Founder	Founder, Chief Executive Officer	President, Senior Vice President
Status	Current	Current	Current	Current

Wells Fargo Success Story

3M Success Story

Platform Overview

Streamlit

Executives

Streamlit Management Team

4 Team Members

You May Also Like

Discover the right solution for your team