Expert.ai demo recording (ENDORSE 2023) - Gianluca Sensidoni
Science & Technology
Introduction
In this demonstration, Gianluca Sensidoni showcases the powerful capabilities of Expert.ai's advanced text analytics and visual analytics tools. The system is designed to enable various features, including clustering and the extraction of specific relations within textual data.
System Features and Navigation
The demo highlights how the platform emphasizes text related to specified devices by highlighting relevant sentences. Users can navigate through vast amounts of data—potentially millions or billions of texts—by clicking on specific nouns to analyze the text in depth. This functionality is particularly useful for examining specific relations between entities.
The system can extract various types of information, including subject-action-object relations and other forms of event-related data. Each sentence is meticulously analyzed, categorizing it by multiple attributes and complements. An important feature is the ability to manage different types of complements reflected within the system.
Entity and Relation Extraction
The platform excels in extracting various entities, ranging from products and devices to geographic locations. Information can be drilled down to specific entity types, such as measures, organizations, and events, utilizing data mapped to geographic databases like geonames.org. The system employs a color coding scheme to distinguish between different categories—these categories may be linked to different taxonomies.
One of the most notable aspects of this tool is its ability to train specific semantic routes with the collaboration of end users and partners. This allows for deeper customization of category extraction based on particular projects or requirements.
Knowledge Graph and Disambiguation
At the core of the technology is a sophisticated knowledge graph comprising approximately two million concepts and about six million relations. This extensive graph allows the system to understand diverse meanings for terms. For example, the word "bank" can represent various concepts—like a financial institution versus the action "to deposit," showcasing different relationships within the ontology.
The system's disambiguation algorithm helps resolve ambiguities in language by navigating the ontology, akin to human comprehension. Just as a person gains insights from reading subsequent pages of a book, the algorithm harnesses previously analyzed data to enhance precision in text analysis.
To achieve this, the tool employs multiple layers of processing, including tokenization, keyword extraction, and grammatical logic, thereby ensuring a precise understanding and representation of the analyzed text.
Decision Support System
All structured information extracted from the analysis can be saved in a decision support system. This feature empowers end users, readers, analysts, and practitioners to make informed decisions based on a comprehensive understanding of the analyzed texts.
Conclusion
In conclusion, Gianluca Sensidoni’s presentation at ENDORSE 2023 demonstrated the versatility and sophistication of Expert.ai's system in navigating and analyzing vast amounts of text data. Interested individuals can reach out for further demonstrations, particularly tailored to specific domains such as life science and telecommunications.
Keywords
- Text analytics
- Visual analytics
- Clustering
- Entity extraction
- Relation extraction
- Disambiguation algorithm
- Knowledge graph
- Decision support system
- Semantic routes
- Taxonomy
FAQ
Q: What is the main functionality of the Expert.ai system showcased in the demo?
A: The main functionality includes advanced text analytics, visual analytics, entity extraction, and relation extraction to analyze large volumes of text efficiently.
Q: How does the system handle disambiguation?
A: The system uses a disambiguation algorithm that navigates a knowledge graph to resolve ambiguities, improving precisions in analyzing text.
Q: Can users customize the system for specific needs?
A: Yes, users can train semantic routes and customize category extraction based on their particular projects or requirements.
Q: What types of entities can be extracted using the system?
A: The system can extract various entity types, including products, devices, measures, organizations, events, and geographic information.
Q: How can the analysis results be utilized?
A: The extracted structured information can be saved in a decision support system, aiding multiple stakeholders in making informed decisions based on the analysis.