Poster image for the video

Train and Deploy Industry‑Leading Generative and Traditional AI.

Data and AI are inextricably linked. We bring 35+ years of data expertise to your AI initiatives. Seven of the world’s largest tech companies trust Innodata for their AI needs.

Poster image for the video
Poster image for the video

Explore Innodata’s Solutions

Generative AI

Develop data to train and refine existing and pre-trained models for task taxonomies. Create large-scale training datasets and golden datasets for supervised fine-tuning.

Rely on human experts-in-the-loop to improve hallucinations and edge-cases with ongoing feedback to achieve optimal model performance.

Address model vulnerabilities, rigorously test and optimize models to ensure safety and compliance, expose model weaknesses, and improve responses to real-world threats.

Collect or create diverse datasets tailored to your needs through natural collection, synthetic generation, studio capture, and on-the-ground collection to train generative AI models.

Coming Soon: Develop advanced LLMs with our GenAI Platform for supervised fine-tuning, prompt writing, response annotation, RLHF model evaluation, and red teaming for rigorous prompt testing to ensure LLM safety.

Traditional AI

Utilize our team of subject matter experts to deliver accurate, secure, reliable, and domain-specific data annotation services across all data types, including image, video, sensor (LiDAR), text, document audio, and speech data.

Collect or create diverse datasets tailored to your needs through natural collection, synthetic generation, studio capture, and on-the-ground collection to train traditional AI models.

Enable your teams to label data at scale with our web-based annotation platform for record classification, document classification, inline classification, and image annotation.

AI for the Enterprise

Align with your business objectives, define automation roadmaps, conduct vision workshops, and find where AI can foster innovation within your organization with Innodata’s AI consultation services. We bring design thinking and strategic planning to establish the right balance of person/machine synergy.

Leverage cutting-edge AI data services for AIOps, CoPilots, and advanced model development with data augmentation, human preference feedback, data for fine-tuning, model configuration and integration, and rigorous model safety, evaluation, and red teaming practices.

Transform your digital infrastructure with our cloud migration, legacy app modernization, and product design services. We help you upgrade outdated systems, modernize legacy apps, and design innovative products to enhance efficiency, performance, and long-term sustainability.

Streamline operations with our workflow automation and managed services. We design automated AI processes tailored to your business, reducing manual tasks and improving efficiency. Our solutions help you scale effortlessly, enabling your teams to focus on innovation and growth.

Utilize our out-of-the-box or industry-specific pre-trained AI models to extract data from any complex document in seconds, enabling powerful automation of document-heavy workflows.

Industry-Specific Platforms

Find journalists and influencers with our media database, monitor media coverage and mentions, measure performance with analytic solutions, and write and release content with generative AI-enabled newswire services.

Extract and analyze medical record data with our AI-enabled platform, allowing you to review medical records with greater efficiency and effectiveness, and create a foundational data layer to enable AI-augmented underwriting.

A Trusted Partner
with Leading
AI Companies

Magnificent Seven Program Manager

Al Research Team

We could not have developed the scale of the current classifiers without Innodata. The average info worker team can’t deal with documents at scale and speed, so Al classifiers for auto-labeling are required; to meet this requirement we needed a huge corpus of sample documents across dozens of business categories.

We went from basically no Al classifiers to 70 Classes in less than a year – from zero to exceeding compete parity. I don’t think there would have been another way to do that without the data corpus. I’m unaware of any other partner than Innodata that could have done that with the speed, volume, accuracy, and flexibility we needed.
Poster image for the video
Years Of Experience
0 +
Languages Supported
0 +
Global Subject Matter Experts
0 +
Customers
0 +

Why Innodata?

STEM and Industry-Specific Domain Expertise

With over 5,000 in-house subject matter experts spanning STEM and industry-specific domains such as healthcare, finance, and legal, along with specialized experts like linguists and taxonomists, Innodata provides industry-leading subject matter expertise for any enterprise-grade AI development.

Quick Turnaround at Scale with Quality Results

Our globally distributed teams guarantee swift delivery of high-quality results 24/7, allowing rapid scalability in local expansion and globalization across projects of any size and complexity.

Global Delivery Centers & Language Capabilities

Innodata operates in 20+ global delivery centers with proficiency in over 85 native languages and dialects, ensuring comprehensive language coverage for your AI projects.

Customized Tooling

Benefit from our proprietary AI SaaS platforms, including our GenAI Workbench (coming soon) for advanced LLM development, our no-code/low-code Annotation Platform for scalable data labeling, and our Document Intelligence for complex data extraction from complex documents.

Poster image for the video

Enabling Domain‑Specific AI Excellence Across Industries

Question + Answering for Global Tech Company

A leading global tech company sought Innodata to enhance their question answering system. This project involved generating accurate and contextually appropriate responses to a wide range of prompts, from straightforward factual questions to complex creative tasks.

  • Diverse Prompt Types: The system needed to handle various prompt types, including factual questions and creative writing tasks.
  • Sensitive Content: Ensuring responses appropriately handled sensitive topics was critical to maintain user trust and safety.
  • Contextual Understanding: Integrating contextual information from chat histories to provide relevant and coherent responses posed a significant challenge.
  • AI Capabilities: Clearly defining the capabilities and limitations of the AI to prevent it from generating responses beyond its scope.

The project adopted a structured approach to address these challenges, focusing on robust training, rigorous evaluation, and continuous improvement.​

Team Preparation:Assembled a team with strong writing skills and experience in generating concise, mobile-friendly content.

Guidelines and Evaluation:​ Developed comprehensive guidelines to ensure responses were accurate, unbiased, and sensitive to user context.Implemented a detailed evaluation process to continuously assess and improve the quality of responses.

Training and Feedback:Conducted extensive training sessions to familiarize the team with the task requirements and the AI’s capabilities.Provided regular feedback to refine the team’s approach and ensure adherence to guidelines.

Contextual Integration:Implemented strategies to effectively utilize chat history and other contextual information to enhance response relevance.

Improved Response Quality: The project significantly enhanced the accuracy and contextual relevance of the AI’s responses.

Increased Efficiency: Streamlined processes and clear guidelines led to quicker and more efficient response generation.

Enhanced User Trust: By effectively handling sensitive content and providing contextually appropriate responses, user trust and satisfaction increased.

Scalability: The refined methodologies and processes proved scalable, enabling the client to handle increasing volumes of user queries effectively.

Intelligent Regulatory Insights with Machine Learning and OpenAI

A large US bank, with a global presence, faced the challenge of staying informed about a constantly evolving landscape of financial regulations. Manually reviewing thousands of regulatory documents published weekly across various sources was a time-consuming and error-prone process. The bank needed a solution to: 

  • Reduce Time Spent on Updates: Legal professionals needed a more efficient way to stay current with regulatory changes. 
  • Improve Information Relevance: Sifting through irrelevant information wasted valuable time and resources. 
  • Ensure Timely Awareness: The bank needed to be promptly informed of any regulatory changes impacting their operations. 
  • Support Informed Decision-Making: Easy access to clear and actionable insights was crucial for informed decision-making. 

Innodata developed an Intelligent Insights program utilizing OpenAI and machine learning to address these challenges. The program leveraged the following:

Comprehensive Regulatory Content Repository: Innodata built a vast repository of regulatory documents obtained through automated scraping techniques. This repository ensured access to the latest and most relevant information. 

Machine Learning for Categorization: Advanced machine learning algorithms were used to categorize the scraped documents based on pre-defined metadata (e.g., jurisdiction, document type). This facilitated efficient information retrieval. 

OpenAI for Intelligent Summaries: OpenAI’s capabilities were harnessed to generate daily or weekly summaries of the categorized regulatory documents. Users could customize their subscriptions based on specific needs (regions, document types) to receive the most relevant information. 

Natural Language Processing (NLP): Leveraging NLP, the system could interpret complex legal language, providing clear and concise summaries in a user-friendly format. 

Real-Time Updates: The system continuously monitored regulatory sources for updates, ensuring users were always informed about the latest changes. 

Source Access: Users could easily access the original PDF source document for any summarized point, facilitating deeper dives when needed. 

The Intelligent Insights pilot program delivered significant improvements:

Increased Efficiency: Legal professionals could stay updated on regulatory changes with minimal time investment, freeing up valuable resources for analysis. 

Enhanced Relevance: Customizable subscriptions ensured users received only the information relevant to their specific needs, eliminating information overload. 

Improved Timeliness: Real-time updates kept users informed about the latest changes as they occurred, minimizing the risk of non-compliance. 

Empowered Decision-Making: Clear and actionable insights from the summaries facilitated informed decision-making within the organization. 

Generative AI Solutions for a Leading Information Publisher

A major information publisher sought to deploy Generative AI within their legal and regulatory operational processes. The goal was to achieve scalability across 13 European countries and multiple languages while maximizing efficiency.

Innodata collaborated with the client to offer specialized consulting and implementation services. The process began with a vision workshop designed to educate stakeholders and identify potential opportunities for AI integration. Two areas were selected for initial 3-month Proofs of Concept (POCs):

Abstract Creation from German Court Cases: Using Generative AI to automate the generation of case summaries. 

Keyword Extraction and Taxonomy Matching for Dutch Labor Law Books: Employing AI to identify relevant keywords and match them to an established taxonomy.

Innodata engaged closely with key stakeholders to establish review and evaluation criteria for each POC. The implementations utilized advanced AI techniques including:

Generative Pre-trained Transformers (GPT): Leveraged for natural language understanding and generation. 

Chain of Density: Applied to ensure coherence and relevance in generated content. 

Prompt Engineering: Used to optimize AI responses. 

Fine-Tuning: Customized the model on specific legal data to enhance performance. 

Vector Database with Similarity Matching: Employed to ensure accurate keyword extraction and taxonomy alignment. 

Innodata demonstrated that Generative AI could enhance processes traditionally reliant on human expertise in language and legal fields. The results underwent a rigorous double-blind review process and were benchmarked against industry standards.

German Abstracts: Of the abstracts generated by GPT, 44% were rated favorably and deemed publishable without modification. In comparison, manually generated abstracts had a 58% approval rate, supporting the efficacy and subjective consistency of the GPT-generated content. 

Dutch Keywords: The GPT system achieved a 25% exact match rate with manually tagged keywords. For reference, a comparison between two human taggers resulted in a 22% exact match, indicating that the GPT solution performs comparably to the existing manual process. 

Both POCs will be further developed for production and scaled across countries based on business needs and priorities. Innodata has also proposed and will facilitate change management strategies to support the program's expansion. Additionally, new opportunities have been identified and transitioned into the POC phase for expedited evaluation.

Image Caption Generation ​

The challenge presented in the Image Caption Generation project lay in crafting detailed and accurate captions for advertisement images, aligning with accessibility standards. Raters were tasked with describing images with precision, catering to visually impaired audiences, while strictly adhering to client guidelines. The task demanded a high level of detail within a limited word count, posing a significant execution challenge. 

To address the challenge, our approach involved thorough training and meticulous preparation. We provided comprehensive learning modules and resources to familiarize the team with the task requirements and client guidelines. A dedicated Q&A system ensured clarity, with trainers promptly addressing any queries or ambiguities. Additionally, we maintained a centralized Source of Truth document for up-to-date guidelines and instructions, streamlining the annotation process. 

Through our tailored training strategy and meticulous approach, we optimized the annotation process, ensuring the generation of high-quality image captions. This not only improved accessibility for visually impaired audiences but also enhanced the overall quality and relevance of advertisement content. By sourcing candidates with relevant experience and conducting rigorous certification assessments, we built a skilled team capable of consistently delivering accurate and descriptive image captions. The project’s success established a scalable framework for future endeavors, further advancing accessibility standards and enhancing the client’s advertisement capabilities.

Streamlining Regulatory Content Management with Automation and Retrieval-Augmented Generation (RAG)

A large US bank, operating across 100+ countries, faced a monumental task in managing a constantly evolving landscape of financial regulations. With thousands of legal documents published weekly across various sources (websites, PDFs, etc.), their legal department struggled with: 

  • Data Volume: Processing thousands of pages of regulations per week manually was time-consuming and inefficient. 
  • Categorization & Retrieval: Manually searching for updates on hundreds of websites was cumbersome and prone to errors. 
  • Time Constraints: Reviewing each source and downloading documents took valuable time away from analysis. 
  • Integration: Manually integrating data into their internal system was a slow and error-prone process. 
  • Scalability: Manual processes couldn’t keep pace with the ever-increasing volume of regulations. 
  • Data Diversity: Regulatory information resided in various formats (PDF, Excel, Word, HTML) across diverse sources. 

Innodata addressed these challenges with a comprehensive approach, leveraging automation and cutting-edge technology:

Automated Workflow Tool: We developed a tool to automate the scraping of regulatory documents from various sources. This ensured a comprehensive and up-to-date dataset for analysis.

Machine Learning for Metadata Enhancement: Machine learning algorithms analyzed the scraped content, automatically assigning relevant metadata like publication date, jurisdiction, and document citations. This improved data organization and searchability.

Following a streamlined data collection process, Innodata built upon the document repository with the core solution:

Retrieval-Augmented Generation (RAG) Tool:  This custom-built tool, designed specifically for financial regulations, uses Azure AI and Cohere’s Coral to answer user queries. It combines retrieval and generation techniques for efficient and insightful responses

  • Azure AI Search: This powerful search engine facilitates efficient document retrieval based on user-entered queries. 
  • Natural Language Processing (NLP): Advanced NLP allows the RAG tool to understand complex legal queries and provide accurate responses derived from retrieved documents. 
  • User-Friendly Interface: Users can easily input queries, view retrieved documents, and understand generated summaries and interpretations presented in a clear and concise format. 

The implemented solutions yielded significant benefits:

Enhanced Efficiency: Manual processes were replaced with automation, significantly reducing the time required to find and interpret regulatory documents. 

Improved Accuracy: Combining Azure AI Search with GenAI ensured highly accurate and relevant responses to user queries. 

Real-Time Updates: The system continuously monitors and indexes new regulations, keeping users informed about the latest changes. 

Scalability: The RAG tool scales effortlessly with increasing data volumes, ensuring a sustainable solution for growing compliance needs. 

Cost Savings: Automation reduced reliance on manual labor and specialized compliance experts, leading to significant cost savings. 

Text Generation in the Advertising Space

A leading FAANG client approached Innodata with the task of enhancing their AI-generated advertising copy. The primary objective was to assess and refine the quality of AI-generated text in comparison to original ads, focusing on content creativity, syntactical and lexical diversity, and the presence of hallucinations. This project aimed to ensure that the AI-generated copy met high standards of creativity and accuracy, vital for maintaining the brand’s reputation and effectiveness in the advertising space. 

The specific challenge stemmed from the limitations of the standard LLM, which relied solely on pre-existing patterns learned during training. This hindered its ability to perform actions requiring specific external information retrieval, hindering its functionality.

Innodata implemented a comprehensive strategy to tackle the challenge. First, we assembled a specialized team with a background in advertising and creative writing, ensuring they had a keen understanding of marketing language and best practices. The team underwent rigorous training, including:

Learning Modules and Practice Exercises: Pre-recorded lessons and exercises to hone creative writing skills and familiarize them with client guidelines.

Q&A Sessions: Regular meetings to address queries and refine understanding, with an emphasis on resolving ambiguities and ensuring adherence to guidelines.

Reverse Engineering: Analyzing client-provided answers to practice tasks to identify patterns and nuances, enabling the team to align closely with client expectations.

Exercise Sets: Creating ad copy, identifying hallucinations, and editing subpar outputs to meet diversity and creativity standards.

By leveraging Innodata's expertise, the FAANG client achieved:​

Enhanced AI Output: Improved quality of AI-generated ads, with a clear distinction between creative and non-creative elements, and diverse yet coherent text.

Brand Integrity: Minimized hallucinations that could harm the brand’s reputation, ensuring that all generated content was accurate and trustworthy.

Operational Efficiency: Established a scalable framework for evaluating and refining AI-generated content, streamlining the process and setting a high standard for future projects.

Client Alignment: Our insights and methodologies often influenced the client’s approach, leading to refinements in their guidelines and enhancing overall project outcomes.

Base Annotations Comparison ​

Innodata partnered with a leading global technology company to improve their AI models, comparing responses from different AI models to determine which performed better based on three quality attributes: helpfulness, honesty, and harmlessness. ​

  • Diverse Prompt Types:  The system needed to assess various types of responses, from factual answers to creative writing. ​
  • Quality Assessment: Responses were evaluated based on their helpfulness, honesty, and harmlessness, requiring nuanced judgment. ​
  • AI Capabilities: Understanding and clearly defining the AI’s capabilities and limitations was essential to accurate assessment. ​

The project focused on training and evaluation strategies to ensure accurate and consistent response comparisons.

Training and Evaluation​:

  • Conducted thorough training to familiarize the team with task requirements and the AI’s capabilities.​
  • Used learning modules, Q&A sessions, and practical exercises to prepare the team.​

Quality Attributes​:

  • Helpfulness: Assessed if the response addressed the prompt coherently, concisely, and relevantly.​
  • Honesty: Evaluated factual accuracy, neutrality, and transparency.​
  • Harmlessness: Ensured responses were safe, non-offensive, and devoid of unqualified advice.​

Response Ranking​:

  • Agents ranked responses based on their adherence to the quality attributes, from best to worst.​​

Innodata significantly improved the AI models of the leading technology company by implementing a structured evaluation process focused on helpfulness, honesty, and harmlessness. Through targeted training and practical exercises, the team efficiently assessed and enhanced the AI's responses. This collaboration not only increased the accuracy and relevance of the outputs but also streamlined the evaluation process, boosting operational efficiency. Innodata's methodologies supported scalable improvements, contributing substantially to the advancement of the company's AI capabilities. 

Enhancing Summarization Accuracy for Compliance​

A leading multinational tech company was in the final stages of developing a cutting-edge language model specifically designed for eDiscovery and communication compliance projects.

A key feature of this model was its summarization skill, crucial for efficiently handling large volumes of documents. However, the tech company faced a significant obstacle: they lacked the necessary documents to rigorously test and validate the summarization capabilities of their model. Without proper testing, the copilot could deliver inaccurate summaries and hinder eDiscovery workflows. 

Innodata’s team of experts created a comprehensive set of 500+ documents tailored for testing the language model’s summarization skills. These documents were carefully curated to reflect the diverse and real-life scenario typically encountered in eDiscovery projects. By providing a robust and varied dataset, we ensured that the model’s summarization feature could be thoroughly tested under realistic conditions. ​

In addition to creating the documents, our team rigorously tested the model using this dataset by developing and implementing a series of prompts designed to push the model’s summarization capabilities to their limits. By simulating various real-world scenarios and document complexities, they were able to thoroughly evaluate and enhance the language model’s performance. These prompts helped categorize the bugs and errors encountered during testing of the created documents into categories such as false positives, false negatives, and graceful responses. This approach resulted in a highly robust model capable of handling complex datasets. 

Bug-Free Deployment: Innodata’s testing helped identify and fix bugs, ensuring the model’s accuracy and reliability. 

Enhanced Summarization Accuracy: Testing led to significant improvements in the copilot’s ability to summarize documents accurately. 

Increased Customer Confidence: With a thoroughly tested copilot, the tech company was confident in its product’s value for eDiscovery professionals. 

Search Summarization ​

A leading tech company approached Innodata with a task requiring the creation of high-quality, user-centric summaries based on search queries and retrieved web documents. The challenge involved: 

  • Aligning with User Intent: Summaries needed to be concise (75-100 words) and directly address the user’s last message within the ongoing conversation. 
  • Accuracy and Originality: Information gleaned from reference documents needed to be presented accurately, with proper citations but avoiding plagiarism. 
  • Adhering to “3 H’s”: Summaries had to be Helpful, High-quality, and Human-rated, ensuring a clear and informative user experience. 

Innodata implemented a comprehensive training program to equip annotators with the necessary skills:​

Learning Resources: Pre-recorded lessons and detailed documentation provided clear guidelines and examples.

Practice Exercises: Interactive Google Forms exercises allowed for practical application of knowledge.

Real-world Experience: Early access to the client’s practice queue ensured familiarity with the actual work interface.

Dedicated Support: Q&A Excel sheets addressed questions and facilitated clarification from the client.

Centralized Knowledge Base: An internal “Source of Truth” document housed the latest guidelines and resources.

Example Sets: Client-provided gold-standard answers served as reference points for annotators.

Skilled Workforce: Recruitment focused on individuals with experience in concise writing (e.g., newsletter writers) and basic fact-checking.

Innodata’s training program ensured the success of this project. The combination of pre-recorded lessons, practical exercises, and real-world application through the practice queue equipped annotators with a strong foundation in the client’s guidelines and best practices. Dedicated support through Q&A channels and a centralized knowledge base addressed any questions or uncertainties, while client-provided gold-standard answers offered valuable benchmarks. Finally, the recruitment strategy focused on individuals with experience in concise writing and fact-checking, perfectly aligning with the project’s demand for accurate and informative summaries.  ​

Ultimately, Innodata’s comprehensive approach not only prepared a skilled workforce for the leading tech company but also helped the client achieve its goal of training an AI model to effectively generate user-centric search summaries. ​

Chatbot Instruction Dataset for RAG Implementation

A leading technology company approached Innodata with a unique challenge. They needed a specialized dataset to train their large language model (LLM) to perform complex “multi-action chaining” tasks. This involved improving the LLM’s ability to not only understand and respond to user queries but also access and retrieve relevant information beyond its initial training data.

The specific challenge stemmed from the limitations of the standard LLM, which relied solely on pre-existing patterns learned during training. This hindered its ability to perform actions requiring specific external information retrieval, hindering its functionality.

Innodata implemented a creative approach to address the client's challenge:

Chain-of-Thought Prompt Development: Innodata’s team of experts employed a technique called “Chain of Thought in Context Learning” to design prompts that encouraged the LLM to explicitly showcase its internal thought process while responding to user queries. This provided valuable insights into the LLM’s reasoning and information retrieval steps.

Prompt Completion with RAG Integration: The team leveraged “Prompt Creation Completion” techniques, where authors set up prompts, craft related queries, and complete the prompts using the Retrieval-Augmented Generation (RAG) tool. This tool retrieved relevant information necessary for the LLM to complete the task at hand.

Author Expertise: Our team of skilled authors, equipped with an understanding of API and RAG dependencies, crafted the dataset elements:

  • User-facing chatbot conversations simulating real-world interactions. 
  • Internal thought processes of the chatbot, revealing its reasoning and information retrieval steps. 
  • System-level instructions guiding the chatbot’s actions. 
  • Training on complex use cases involving multi-step tasks and subtasks. 

The resulting dataset, enriched with the "chain-of-thought" approach, offered the client significant benefits:

Enhanced LLM Functionality: The dataset equipped the LLM with the ability to perform complex, multi-action tasks, significantly improving its practical applications.

Improved Information Retrieval:  By incorporating the RAG tool, the LLM gained the ability to access and retrieve crucial information from external sources, overcoming its prior limitations.

Deeper Model Understanding: The “chain-of-thought” element provided valuable insights into the LLM’s reasoning process, enabling further optimization and development.

Creating Health and Medical Dialogues Across 8+ Specialties

A leading medical publisher approached Innodata with a critical need. They required a comprehensive dataset of medical dialogues, spanning over 8 different specialties, to support advancements in medical knowledge retrieval and automation. This dataset would serve as the foundation for semantic enrichment – a process that enhances the understanding of medical information by computers.

The key requirements were:

  • Multi-Specialty Focus: Dialogues needed to cover a wide range of medical sub-specialties, exceeding 20 in total. 
  • Real-World Tone: The dialogues should mimic genuine conversations within medical settings, while referencing the client’s specific “clinical key” as a knowledge base.
  • Pre-Determined Topics: The client provided a list of medical and health areas to ensure the dialogues addressed relevant issues.
  • Exceptional Accuracy: Achieving 99% accuracy in the medical content of the conversations was paramount.

Innodata implemented a multi-step workflow to deliver a high-quality medical dialogue dataset:

Expert Actor Recruitment: Innodata assembled a team of actors with real-world medical experience, including nurses, medical doctors, and students. This ensured the dialogues reflected the appropriate level of expertise and communication style for each scenario. 

Content Development: Our medical writers crafted the dialogues based on the client’s provided topics and “clinical key” resources. Each conversation maintained a natural flow while adhering to strict medical accuracy.

Multi-Layer Review: The dialogues underwent a rigorous review process by medical professionals to guarantee factual correctness and adherence to the 99% accuracy benchmark.

By leveraging Innodata's expertise in medical content creation and actor recruitment, the client received a unique and valuable dataset:

Extensive Medical Coverage: The dataset encompassed dialogues across a broad spectrum of medical specialties, providing a robust foundation for various applications.. 

Realistic Interactions: The diverse cast of actors and natural dialogue style ensured the dataset accurately reflected real-world medical communication.

Highly Accurate Content: The 99% accuracy level guaranteed the dataset’s suitability for training AI models and enriching medical knowledge retrieval systems.