Faster then a human would do the same job
Different types of extraction including tables
We are saving 6 months of manual work
The German company LECTURA provides information about heavy equipment: agricultural machinery, cranes, construction machinery, etc. They have tens of thousands of items listed on their website, and each of them has a downloadable PDF of technical specifications available.
The company wanted to achieve better SEO results and offer more accurate results in a narrowly defined area to niche down and differentiate itself from the competition. However, such a huge volume of technical specification text is impossible to process manually, so they turned to us. Our goal was to paraphrase the English catalogue texts to make them unique which results in better rankings in search engines. In this way, we help customers in a highly specialised field find the exact results they are looking for.
To do this, we used our paraphrasing tool, which you can also try out in our demo section. It also rephrases text to better reflect the terminology people use when searching.
„We can now use algorithms to rewrite content into more domain-specific language,“ adds Michal Štefánik, our data specialist with a focus on natural language processing and machine learning.
For paraphrasing, we use a generative language model trained on a large corpus of texts. We first train it on specific terminology using client data, and then feed it with keywords that users most often go to the product page through. Finally, we set the model to use these selected keywords when paraphrasing. This makes the text rank better in the search engine.
It is no longer the case that AI is only applicable to a small number of standard use cases – and our collaboration is proof of that. We have the expertise to unlock the potential of AI across a wide range of applications.
LECTURA is the leading provider of machinery intelligence supporting over 1M visitors monthly.