Deep learning, aka Artificial Intelligence, aka Neural Networks, is everywhere now: from automatic translation, to text generation with GPT-3.
But search is also affected by this trend. A new generation of so-called “Vector search” libraries or services are emerging. But the missing link is missing to make them widely adopted: simplicity and low cost.
WPSOLR intends to be this missing link, by bringing those extraordinary features to WordPress and WooCommerce, as it did for Elasticsearch, Solr and Algolia.
Some of the neural search, neural similarity, open-source or managed services, that WPSOLR is currently investigating:
Some of the features we would like to bring to WordPress:
Latest Models: Use all the latest transformer-based models (eg, BERT, RoBERTa, MiniLM) for extractive quality assurance, generative quality assurance, and document retrieval.
Modular: multiple choices to fit your technology stack and use case. Choose your favorite database, file converter, or modeling framework.
Open: 100% compatible with the HuggingFace model hub. Close interfaces with other frameworks (e.g. Transformers, FARM, sentence transformers)
Scalable: Scale to millions of documents through retrievers, production ready backends like Elasticsearch / FAISS, Google latests ANN algorithms, and top-notch Vector databases.
End-to-end: all tools in one place: file conversion, cleanup, split, training, evaluation, inference, labeling, etc.
Developer friendly: easy to debug, extend and modify.
Customizable: adjust the templates to your domain or implement your custom DocumentStore.
Continuous learning: Collect new training data from user feedback in production and improve your models
And because a search needs real data, we already made connectors to the most used plugins.
Work seamlessly with WooCommerce, bbPress, ACF PRO, Toolset Views, WPML, Polylang, Yoast SEO, and many more.
We can also build one for you if necessary.
Studio WPSOLR is the team behind WPSOLR, our flagship search plugin for WordPress & WooCommerce.
The Studio cumulates years of experience in development and configuration for:
Our WPSOLR search plugin is probably the most flexible and scalable on the market. It fits 99% of our customers needs.
But sometimes, one needs a bit of extra hand to reach the perfect experience.
We can help you:
Another MyListing client came to us because his site was unable to cope with his million listings.
Not only the search was slow, but also the home page and the Yoast sitemaps.
We modified our MyListing add-on, and now the client site is flying, even with one million listings.
A MyListing client came to our Studio because he couldn’t find listings with enough accuracy: some custom fields where not searched at all.
We had to build a full add-on, because myListing is not using the standard archive search.
Now, with the new add-on, our client visitors can search in every bit of information related to the listings.
The customer took 4,000 images of the Grocery’s store products.
The products are imported (synchronized) from a Kezia II cash register software, into the WooCommerce shop “Magasin Bio à La Teste de Buch“.
Images are then uploaded in the media library.
Question:
How one can match the images to the product titles, so to upload the correct image to the product images gallery?
Answer:
By extracting the image texts (OCR), using the WPSOLR add-on for Google Vision API.
Then by indexing the mages in Elasticsearch with their OCR texts.
After that, opening the media library and looking for “Wine Bordeaux 2019” returns the images that need to be attached to the related product.
A client came to us because he noticed that his search was often ranking old posts in first positions. Which is not so good for a newspaper.
On the other hand, sorting by date was showing irrelevant posts in first positions.
So, we built a new add-on based on a decay feature of Elasticsearch. This feature enables to rank results according to both their accuracy and their dates. Even better, the ratio accuracy/date can be fine tuned in the WPSOLR admin.
English▼
