---
title: The Science Behind Tellagence — Published AI Research
description: Two arXiv papers with Villanova and Portland State show why AI sentiment tools disagree with themselves, and the framework that fixes it.
image: https://tellagence.ai/hubfs/site-images/tellagence-share-1200x630.png
---

[Tellagence](https://tellagence.ai/?hsLang=en)

Menu

[Case Studies](https://tellagence.ai/case-studies?hsLang=en)[See the Science](https://tellagence.ai/science?hsLang=en)[Pricing](https://tellagence.ai/pricing?hsLang=en)[About Us](https://tellagence.ai/about?hsLang=en)[Get a demo](https://tellagence.ai/science#demo)

See the science

# The science behind the solution.

Tellagence is consumer intelligence software built on years of original research, published on arXiv with Villanova University and Portland State University. It solves one problem: making what people say in public reliable enough to decide on.

[Read the research](https://tellagence.ai/science#research) [Meet Nitin Mayande, CSO](https://tellagence.ai/science#nitin)

Most AI tools are wrappers. They take an existing large language model and build a layer on top. These give you an answer, but face some fundamental limitations: unpredictable outputs, poor performance on chaotic real-world data, and results that change every time you run the same query.

At Tellagence, we took a different path. Starting in 2011, Nitin Mayande and his team set out to build something the field didn’t yet have: a deterministic contextual intelligence framework. One that doesn’t have to guess, and can give you the same answer twice.

Up to 30% improvement in data quality. Verified across three independent datasets.

## The research

Two research papers, published on arXiv with Villanova University and Portland State University. Together they measure the reliability problem in AI text analysis and prove the framework that fixes it.

Your AI isn't confused. It just never had the right context.

![Mathematical formulas and curves glowing over a dark grid](https://tellagence.ai/hs-fs/hubfs/site-images/AdobeStock-1526188381.webp?width=1200&height=633&name=AdobeStock-1526188381.webp)

### Leveraging Weighted Syntactic and Semantic Context Assessment Summary (wSSAS) Towards Text Categorization Using LLMs

**Published:** arXiv, April 13, 2026  
**Authors:** Shreeya Verma Kathuria, Nitin Mayande, Sharookh Daruwalla (Tellagence Inc.), Nitin Joglekar (Villanova University), Charles Weber (Portland State University)

A new deterministic framework for contextual intelligence that fixes the core reliability problem with AI-powered text analysis, and proves it works. Up to 30% improvement in data quality, verified across three independent datasets.

[Read the explainer](https://tellagence.ai/pulse-feed/ai-text-categorization-explained?hsLang=en) [Download the full paper](https://tellagence.ai/hubfs/reports/Leveraging-wSSAS-Towards-Text-Categorization-Using-LLMs.pdf?hsLang=en)

Ask your sentiment tool the same question twice. We'll wait.

![Illustrated brain surrounded by equations and glowing data curves](https://tellagence.ai/hs-fs/hubfs/site-images/AdobeStock-506710755.webp?width=1000&height=500&name=AdobeStock-506710755.webp)

### Consistency Analysis of Sentiment Predictions Using Syntactic & Semantic Context Assessment Summarization (SSAS)

**Published:** arXiv, April 16, 2026  
**Authors:** Sharookh Daruwalla, Nitin Mayande, Shreeya Verma Kathuria (Tellagence Inc.), Nitin Joglekar (Villanova University), Charles Weber (Portland State University)

The AI sentiment analysis you’ve been using gives different answers every time. Our contextual intelligence research measured the problem. Then we solved it: a deterministic framework returns the same answer on the same data every time it runs.

[Read the explainer](https://tellagence.ai/pulse-feed/ai-sentiment-analysis-consistency?hsLang=en) [Download the full paper](https://tellagence.ai/hubfs/reports/Consistency-Analysis-of-Sentiment-Predictions-Using-Syntactic-Semantic-Context-Assessment-Summarizat.pdf?hsLang=en)

![Nitin Mayande headshot](https://tellagence.ai/hs-fs/hubfs/site-images/Nitin-Headshot.webp?width=600&height=900&name=Nitin-Headshot.webp)

## Nitin Mayande

Co-Founder and CSO, Tellagence

Nitin Mayande has spent more than 15 years building AI that reads language the way people do. A Ph.D.-trained engineer, he holds multiple patents on AI systems and has worked on large-scale data, machine learning, and contextual understanding, including at Nike and Intel.

The research on this page represents just part of an ongoing body of scientific work that forms the foundation of Tellagence’s contextual intelligence framework.

## Read the explainer posts on our blog

[![Your Sentiment Tool is Guessing. Here’s the Math.](https://tellagence.ai/hubfs/blog-images/post-410354222800.jpg)](https://tellagence.ai/pulse-feed/ai-sentiment-analysis-consistency?hsLang=en)

### [Your Sentiment Tool is Guessing. Here’s the Math.](https://tellagence.ai/pulse-feed/ai-sentiment-analysis-consistency?hsLang=en)

by Nitin Mayande, Co-Founder and Chief Science Officer, Tellagence

[Read more →](https://tellagence.ai/pulse-feed/ai-sentiment-analysis-consistency?hsLang=en)

[![Why AI Gets Confused — And How We Fixed It](https://tellagence.ai/hubfs/blog-images/post-410354222802.webp)](https://tellagence.ai/pulse-feed/ai-text-categorization-explained?hsLang=en)

### [Why AI Gets Confused — And How We Fixed It](https://tellagence.ai/pulse-feed/ai-text-categorization-explained?hsLang=en)

by Nitin Mayande, Co-Founder and Chief Science Officer, Tellagence

[Read more →](https://tellagence.ai/pulse-feed/ai-text-categorization-explained?hsLang=en)

FAQ

## Questions about the research

Why do AI sentiment tools give different answers to the same question?

Most AI tools are wrappers around a large language model, and those models are not deterministic: run the same query twice and the output can change. Our research measured that inconsistency.

What is deterministic contextual intelligence?

It is the framework behind Tellagence. It assesses the syntactic and semantic context of each piece of text before classifying it, so the same data returns the same answer every time it runs.

How much does it improve results?

Up to a 30% improvement in data quality, verified across three independent datasets.

Who did the research?

Tellagence researchers Shreeya Verma Kathuria, Nitin Mayande and Sharookh Daruwalla, with Nitin Joglekar of Villanova University and Charles Weber of Portland State University. Both papers were published on arXiv in April 2026.

## Want to see what consumer intelligence can find in your data?

The research is the foundation. Let us show you how the Tellagence tool can analyze your language-based data to reveal insights and strategic opportunities like never before.

Work email

Get a demo

Tellagence

Consumer intelligence software.  
Portland, Oregon.  
SOC 2 Type II compliant.

© 2026 Tellagence, Inc. All rights reserved.

[Case Studies](https://tellagence.ai/case-studies?hsLang=en) [See the Science](https://tellagence.ai/science?hsLang=en) [Blog](https://tellagence.ai/pulse-feed?hsLang=en) [Pricing](https://tellagence.ai/pricing?hsLang=en) [About Us](https://tellagence.ai/about?hsLang=en) [Get a Demo](https://tellagence.ai/science#demo) [Contact](https://tellagence.ai/contact?hsLang=en) [Privacy Policy](https://tellagence.ai/privacy-policy?hsLang=en) [LinkedIn](https://www.linkedin.com/company/tellagence/)

```json
{
  "@context" : "https://schema.org",
  "@id" : "https://tellagence.ai/#org",
  "@type" : "Organization",
  "address" : {
    "@type" : "PostalAddress",
    "addressCountry" : "US",
    "addressLocality" : "Portland",
    "addressRegion" : "OR"
  },
  "description" : "Tellagence is consumer intelligence software that reads public conversation in context and returns the same answer every time.",
  "founder" : [ {
    "@type" : "Person",
    "name" : "Matt Hixson"
  }, {
    "@type" : "Person",
    "name" : "Nitin Mayande"
  } ],
  "foundingDate" : "2011",
  "legalName" : "Tellagence, Inc.",
  "name" : "Tellagence",
  "sameAs" : [ "https://www.linkedin.com/company/tellagence/" ],
  "url" : "https://tellagence.ai"
}
```

```json
{
  "@context" : "https://schema.org",
  "@type" : "ScholarlyArticle",
  "author" : [ {
    "@type" : "Person",
    "name" : "Shreeya Verma Kathuria"
  }, {
    "@type" : "Person",
    "name" : "Nitin Mayande"
  }, {
    "@type" : "Person",
    "name" : "Sharookh Daruwalla"
  }, {
    "@type" : "Person",
    "name" : "Nitin Joglekar"
  }, {
    "@type" : "Person",
    "name" : "Charles Weber"
  } ],
  "datePublished" : "2026-04-13",
  "headline" : "Leveraging Weighted Syntactic and Semantic Context Assessment Summary (wSSAS) Towards Text Categorization Using LLMs",
  "publisher" : {
    "@type" : "Organization",
    "name" : "arXiv"
  },
  "url" : "https://245728351.fs1.hubspotusercontent-na2.net/hubfs/245728351/reports/Leveraging-wSSAS-Towards-Text-Categorization-Using-LLMs.pdf"
}
```

```json
{
  "@context" : "https://schema.org",
  "@type" : "ScholarlyArticle",
  "author" : [ {
    "@type" : "Person",
    "name" : "Sharookh Daruwalla"
  }, {
    "@type" : "Person",
    "name" : "Nitin Mayande"
  }, {
    "@type" : "Person",
    "name" : "Shreeya Verma Kathuria"
  }, {
    "@type" : "Person",
    "name" : "Nitin Joglekar"
  }, {
    "@type" : "Person",
    "name" : "Charles Weber"
  } ],
  "datePublished" : "2026-04-16",
  "headline" : "Consistency Analysis of Sentiment Predictions Using Syntactic & Semantic Context Assessment Summarization (SSAS)",
  "publisher" : {
    "@type" : "Organization",
    "name" : "arXiv"
  },
  "url" : "https://245728351.fs1.hubspotusercontent-na2.net/hubfs/245728351/reports/Consistency-Analysis-of-Sentiment-Predictions-Using-Syntactic-Semantic-Context-Assessment-Summarizat.pdf"
}
```

```json
{
  "@context" : "https://schema.org",
  "@type" : "Person",
  "alumniOf" : "Portland State University",
  "jobTitle" : "Co-founder and Chief Scientific Officer",
  "name" : "Nitin Mayande",
  "worksFor" : {
    "@id" : "https://tellagence.ai/#org"
  }
}
```

```json
{
  "@context" : "https://schema.org",
  "@type" : "FAQPage",
  "mainEntity" : [ {
    "@type" : "Question",
    "acceptedAnswer" : {
      "@type" : "Answer",
      "text" : "Most AI tools are wrappers around a large language model, and those models are not deterministic: run the same query twice and the output can change. Our research measured that inconsistency."
    },
    "name" : "Why do AI sentiment tools give different answers to the same question?"
  }, {
    "@type" : "Question",
    "acceptedAnswer" : {
      "@type" : "Answer",
      "text" : "It is the framework behind Tellagence. It assesses the syntactic and semantic context of each piece of text before classifying it, so the same data returns the same answer every time it runs."
    },
    "name" : "What is deterministic contextual intelligence?"
  }, {
    "@type" : "Question",
    "acceptedAnswer" : {
      "@type" : "Answer",
      "text" : "Up to a 30% improvement in data quality, verified across three independent datasets."
    },
    "name" : "How much does it improve results?"
  }, {
    "@type" : "Question",
    "acceptedAnswer" : {
      "@type" : "Answer",
      "text" : "Tellagence researchers Shreeya Verma Kathuria, Nitin Mayande and Sharookh Daruwalla, with Nitin Joglekar of Villanova University and Charles Weber of Portland State University. Both papers were published on arXiv in April 2026."
    },
    "name" : "Who did the research?"
  } ]
}
```