Mindgeist implements a voice interaction solution for libraries, solving technical challenges in DAISY format management and semantic search and significantly improving access to content.
Key points
- Management of DAISY formats for playback on voice assistants.
- Semantic search that improves accuracy in complex queries.
- Solution adapted to the limitations of the MARC format in bibliographic databases.
From Precision to Meaning
Searching for bibliographic content in systems based on MARC databases runs into inherent limitations because it relies on exact matches or rigid patterns. Terms such as “1984” or “Las Mil y Una Noches” (One Thousand and One Nights) create ambiguity when interpreted as numbers, dates or titles of works, and variations such as “El Quijote” do not correspond exactly to the full title of Don Quijote de la Mancha.
Mindgeist designed a solution that replaces the reliance on precise matches with a semantic search able to understand meanings and contexts in order to resolve these ambiguities.
Solution Implemented
Design of a conversational model
Slot-based systems face the challenge of interpreting numerical or textual values:
- Dynamic Configuration:
- Slots adjust to the type of query, processing numbers, dates and titles without confusion.
- Correction of Variations:
- Correction modules convert phonetic or partial descriptions into processable terms.
Semantic Search
We implemented a search that goes beyond exact matching to focus on meaning:
- Query Processing:
- Resolves and normalizes ambiguous inputs (e.g., “mil novecientos ochenta y cuatro” → “1984”).
- Recognizes common aliases or non-literal descriptions, such as “El Quijote”.
- Context-Based Ranking:
- Answers are ordered by semantic relevance, prioritizing titles that match in meaning over partial matches.
- As a side effect, it allows works to be discovered from their description and review.
DAISY Format Conversion
The DAISY format enables advanced accessibility, but it presents technical challenges for playback on voice assistants:
- Conversion to a Compatible Format:
- An automatic pipeline converts DAISY files to formats accepted by Alexa, preserving bookmarks, sections and key points.
- Compatibility Validation:
-
Ensures the metadata is readable and useful for voice searches.
-
Adaptation of MARC Databases
The MARC databases were extended with semantic search layers:
- Semantic Enrichment:
- We process additional information to improve the answers, such as common aliases and regional variants.
Impact of the Solution
Key Results:
- Search Accuracy:
- Semantic search reduced errors in unresolved queries.
- Greater Inclusion:
- Users with different levels of technological experience can access the content intuitively.
- Automation and Scalability:
- Automatic processes in the DAISY conversion made it possible to quickly expand the available collection.
- Resolution of Ambiguities:
- Ambiguous queries such as “1984” or “El Quijote” now return relevant results.
Beyond the Exact Match
Mindgeist’s experience in semantic search and conversational systems made it possible to turn a technical challenge into a functional, accessible solution. This case shows that understanding meaning, beyond the literal text, is key to improving voice interaction in complex environments.
Want to explore how these solutions can be integrated into your project? Contact us to discuss your needs.
Looking to implement voice interaction technologies in your projects? Mindgeist has the experience to take your vision to the next level.
Learn More
Explore the Accessible Library skill directly on Alexa and discover how voice interaction transforms accessibility: