Lexical, Vector & Hybrid Search with Elasticsearch

Item: Lexical, Vector & Hybrid Search with Elasticsearch
Rating: 4
Author: PostSingularity Institute

by Carly Richmond

Applied Sciences & Engineering Computing & Cybersecurity UNDatabases UNHInformation retrieval

🎙 Carly Richmond 👥 1.1M 📅 December 17, 2025 ⏱ 36 min 👁 1K 📄 expert opinion 🧭 2026-08-02

Available in: English (current) Français

Keywords

BM25inverted indexembeddingshybrid searchRAG

Summary

Carly Richmond, a developer advocate at Elastic, presents an overview of search techniques, contrasting traditional lexical search with modern vector search and their combination into hybrid search. She begins by tracing the history of search engines, from Archie to modern distributed systems like Elasticsearch, explaining the inverted index and tokenization process. She details the Okapi BM25 algorithm used for lexical ranking and highlights its limitations, such as vocabulary mismatch and inability to capture semantics. She then introduces vector search, where documents and queries are converted into embeddings, enabling semantic similarity. The talk covers how hybrid search combines both approaches to improve relevance, and briefly touches on reranking models. Finally, she discusses the role of AI and RAG (Retrieval-Augmented Generation) in grounding LLMs with private data, emphasizing the importance of search in the AI era. The presentation includes practical examples and audience interaction, making it accessible yet informative for developers.

149 words

Critical Evaluation

The talk provides a solid, high-level overview of search technologies, effectively contrasting lexical and vector approaches. The speaker demonstrates deep expertise, explaining complex concepts like BM25 and embeddings in an accessible manner. The historical context and practical examples (e.g., Star Wars queries) help illustrate the limitations of lexical search. The argumentation is coherent, building from foundational concepts to advanced hybrid techniques. However, the talk is more of an expert opinion than a rigorous scientific presentation; it lacks formal citations and empirical data to support claims. The discussion of hybrid search and reranking is brief, and the section on AI search could delve deeper into implementation details. The title accurately reflects the content, and the talk is well-structured. The presence of a sponsor segment is noted but does not detract from the technical value. Overall, the information is reliable and aligns with industry knowledge, though it would benefit from more in-depth analysis and references.

153 words

Title / Content Match

The title accurately reflects the content, which covers lexical, vector, and hybrid search approaches using Elasticsearch.

Quality & Reliability

8/10

The talk is presented by a recognized expert from Elastic, with clear technical explanations and references to established algorithms (BM25) and concepts. The content is well-structured and aligns with industry knowledge, though it lacks formal citations and is based on the speaker's experience.

Chapters

Cited Sources

Carly Richmond's Bluesky profile — Speaker's social media profile for further engagement.
GOTO Conferences Bluesky profile — Conference organizer's social media profile.
Carly Richmond's Medium blog — Speaker's blog with additional articles on search and technology.
Carly Richmond's personal website — Speaker's personal site with links to talks and resources.
Carly Richmond's GitHub — Speaker's GitHub repository for code examples and projects.
GOTO Copenhagen conference page — Conference website for event details.
Session page for this talk — Official session page with slides and abstract.
GOTO Conferences LinkedIn — Conference organizer's LinkedIn page.
Carly Richmond's LinkedIn — Speaker's LinkedIn profile.
GOTO Conferences YouTube channel — Conference channel for more talks.

Concurring Sources

Okapi BM25 - Wikipedia — Confirms the algorithm's details and history.
Inverted index - Wikipedia — Supports the explanation of how lexical search indexes documents.
Word embedding - Wikipedia — Provides background on how vector representations are created.

External References

Contribution & Novelties

The talk provides a clear, practical overview of search techniques, emphasizing the importance of hybrid search for modern applications. It bridges the gap between traditional lexical search and AI-driven semantic search, offering actionable insights for developers. The speaker’s experience at Elastic adds credibility, and the examples illustrate real-world challenges.

Pour aller plus loin :

Okapi BM25 — The ranking algorithm discussed, with detailed explanation.
Inverted index — Core data structure for lexical search.
Word embedding — Foundation of vector search.
Retrieval-Augmented Generation — How search grounds LLMs.
Elasticsearch documentation — Official docs for implementing these techniques.

96 words

Radar Profile

The radar profile shows high scores in information quantity and quality, indicating a content-rich talk with reliable information. The technical level is moderate, suitable for a broad developer audience. The overall reliability is strong, reflecting the speaker's expertise and alignment with established concepts.

Reliability 8/10