What is Engineering Village AI?

Engineering Village AI (artificial intelligence) leverages semantic search technology to enable computers to understand the intent and context behind a user’s query. Using this technology, Engineering Village AI analyzes these queries to locate relevant documents in the search index. These identified documents (published in Compendex since 2003) are leveraged to generate summaries and other features using generative AI techniques. To maintain transparency, Engineering Village AI follows specific guidelines to ensure that the summaries are based upon accurate and pertinent Compendex content and include references.

Engineering Village AI was developed using technologies from Elsevier and other sources. It leverages internal and external large language models (LLMs) such as OpenAI's Generative Pre-Trained Transformer (GPT). The Engineering Village AI uses the OpenAI GPT in a private and secure manner, as there is no data exchange or use of the Compendex data to train this LLM.

1. You enter a natural language query, term, or phrase into the EV AI text box and click ‘Search.’
2. Your query is processed using natural language processing (NLP) techniques to extract pertinent information from the text.
3. Your processed query is analyzed to understand the user's intent, context, and information needs. This may involve identifying keywords and relationships within the text.
4. Your query is mapped to a high-dimensional vector representation, also known as an embedding, which encodes its semantic meaning.
5. Based on your query, the system identifies the most relevant documents from the vector search index. Cosine similarity is used to calculate the similarity between the query vector and the vectors of the documents from the index.
6. The documents with the highest similarity scores to the query vector are retrieved.
7. The documents (also known as search results) are ranked based on the similarity scores of the query vector and the documents, with the most similar items ranked as having the highest relevance to your query. The results are sorted to present the most relevant information first.
8. Your query, the document dataset, and the specific instructions are forwarded to the large language model (LLM).
9. The LLM is requested to generate four similar queries.
10. The system uses the four LLM-generated queries and your user query (five queries total) to get related documents.

    • This process produces 20 documents per query, resulting in 100 documents.
11. The 100 documents generated from these queries are reranked and the top 20 are selected.
12. These top 20 documents with queries are passed to the LLM.
13. The LLM processes this information to generate responses, such as the summary or follow-up questions.
14. The search results to your original query are displayed as an expanded summary. The corresponding document references for this summary are also accessible from the footnotes or the ‘View all references’ button beneath the summary.
15. For further research on your topic of interest, you may click any of the follow-up questions (under the ‘Go deeper’ label) that have been provided to you.

In natural language processing, a vector can represent various data points (such as words and phrases). Vectors capture the numerical representations of these data points in a high-dimensional space. Similarity between data points is determined by measuring the distance or similarity between their corresponding vectors in space. The aim is to retrieve data points that are the most similar to a specific query vector. Engineering Village AI uses language models to assign a numerical value to the words in a query based on a language vector space. These numbers are compared to the vectors that have been assigned to the Compendex documents published since 2003, and then choose those that are the best match. These documents are used to generate the summaries, document references, and corresponding follow-up questions.

Engineering Village AI follows GDPR (General Data Protection Regulation) closely to ensure user privacy and limits data retention to only what is necessary. Personal user information and chat history are not stored on the Engineering Village system unless accomplished in compliance to improve the product. Engineering Village is committed to ‘RELX Responsible AI Principles,’ which include promoting accountability, enforcing strong information management, and eliminating bias.

Although Engineering Village AI relies on trustworthy Compendex data content, there is a possibility of receiving inaccurate, distorted, or objectionable results. Engineering Village AI does not offer engineering design, medical, financial, legal, or safety advice. Apply caution before including its results in official documents like dissertations or research papers. Avoid entering personal, confidential, or sensitive information to protect the privacy and integrity of your research.

When using AI-generated content, it is essential to perform independent research. Additionally, Engineering Village recommends you review the guidelines established by your workplace or institution before incorporating content generated by AI.

Refer to ‘How do I use the Engineering Village AI feature?’ for tips and best practices for using this search feature.