How Conveyer Uses AI to Unlock the Power of Product Documentation

Context-driven NLP is the key to turning manuals and other technical documentation into dynamic digital resources.

At Conveyer, we’re in the business of making product documentation more easily accessible. Instead of leaving vital knowledge locked away in hard-to-read paper or PDF product manuals, we use powerful AI tools to quickly and efficiently turn existing technical documentation into a dynamic, mobile-friendly digital resource. 

The results speak for themselves: digitized documentation elevates the customer experience by putting authoritative product information at users’ fingertips in the exact moment that it’s needed. Customized digital solutions also empower businesses to foster enduring post-sale relationships, unlock new revenue opportunities, and capture transformative data insights. 

But the question remains: how exactly does Conveyer convert static paper or PDF manuals and other technical resources into dependable, intuitive digital experiences?

The answer lies in Conveyer’s AI Conversion Engine (ACE), a powerful solution that leverages proprietary rule-based AI algorithms, large language models, and natural language processing innovations to organize, paraphrase, and conversationalize existing documentation, delivering game-changing results with little or no effort from your team. 

Want to know more? Read on to discover three ways that Conveyer’s AI tools take your technical documentation to the next level. 

1. Contextual analysis surfaces key topics

Most product documentation consists of unstructured data. That might sound counterintuitive: your manual is probably broken into sections to guide the reader, for instance. From a data processing standpoint, though, such divisions are invisible; the manual is just a single wall of undifferentiated text. 

As anyone who’s tried scrolling through pages of text on their smartphone knows, that isn’t an effective way to deliver digital content. To create mobile-friendly digital assets, you need to create structure and leverage it to make your content discoverable and accessible. Crucially, you need to do so without in any way changing or rewriting the content already painstakingly developed and vetted by your team. 

That’s where Conveyer’s proprietary AI tools — the result of years of research and refinement — come in. Let’s say you have an eight-page dishwasher user guide: Conveyer’s ACE uses markers already present in your documentation — punctuation, white space, section headings, indexes and table of contents, and more — to organize your manual into discrete idea blocks called Topics. There’s no distortion of your existing documentation — just the rapid distillation of eight pages of unstructured data into clearly defined units, each dealing with a specific concept or idea. 

The result: digital resources that are as authoritative as the original documentation, but that have the structure required to deliver a compelling online experience. Each Topic is hosted within a virtual document that can be linked to, displayed, searched, or indexed independently. Text, images, video, and other content can be delivered seamlessly, and access can be streamlined using QR codes and digital search. Best of all, engagement can be tracked by Topic to unlock deep and granular insights into the user experience.

Topics make FAQs more effective

Many consumers prefer to scan FAQs instead of reading the full manual — and while product manuals usually contain the answers to just about any question a user could have, manually curating an exhaustive list of question-and-answer pairs is costly and time-consuming. 

That’s where Conveyer’s ability to automatically distill undifferentiated content into structured Topics becomes incredibly powerful. By leveraging Large Language Models (LLMs), it’s possible to quickly paraphrase any individual Topic into unique question-and-answer pairings that can be delivered via your product’s FAQ list.

Of course, you could feed your entire manual into any LLM tool and have it spit out a series of FAQs — but the results might not be very useful. The paradox of LLMs is that they produce answers that sound plausible, but that aren’t necessarily factually accurate — the exact opposite of what most users are looking for when perusing a list of FAQs. 

Conveyer solves that problem by narrowing the scope of the text on which an LLM operates. Instead of letting an AI tool loose on your entire manual, we use Topics to right-size each prompt before submitting it for paraphrasing. We then validate Q&A pairs against the source Topic, ranking for both salience and quality to ensure that the results are useful and accurate. And because each Q&A pairing derives from a specific Topic, Conveyer can proactively surface salient FAQs based on the specific needs and search patterns of any given user. 

3. FAQs elevate real-time conversations

Chatbot-style AI tools are increasingly part of the zeitgeist, with ChatGPT now having over 100 million monthly users, and Microsoft even adding AI chat to its Bing search engine. For product support queries, organic chat-style interactions are the natural next frontier.

Here, again, Conveyer’s structured Topics pay important dividends. Once unstructured data is organized into Topics, and enhanced through the addition of paraphrased question-answer pairs, it becomes far easier for conversational AI tools to accurately answer product queries. 

Here’s how it works: Conveyer uses Natural Language ProcessingNLP) to parse consumer questions and compare them to existing Q&A pairings extracted from individual Topics. By categorizing the end-users’ intent — based on what they write or say — we can match their expression to the best possible Q&A pairing. 

Again, the key benefit here is that Conveyer anchors the responses generated by your chatbot in the organizational logic already present in your user manuals. Instead of giving a black-box AI free reign to produce responses that merely sound accurate, we align responses around Topics directly derived from your existing materials, delivering the power of chat-based interactions without sacrificing saliency or accuracy. 

👑Context is king

Because Conveyer’s AI works by contextually deriving structure from existing documentation, it sidesteps the thorniest issues surrounding conventional NLP solutions. 

Today’s top LLMs, including GPT-3, require vast amounts of public data to train their algorithms, enabling them to interpret novel texts and generate human-sounding responses. But precisely because LLMs are trained on public data, they provide generic responses instead of dependable model-specific information. 

The obvious solution is to use product documentation to train NLP tools — but a single unstructured product manual doesn’t typically offer enough text to effectively train an algorithm. Technical documentation also offers only a single perspective — the brand’s — and doesn’t provide trainable insights into queries that customers and end-users have in the real world.

Conveyer’s context-first approach solves both problems. By sorting documentation into Topics, Conveyer’s ACE distinguishes the signal from the noise to support model-specific AI and NLP solutions, including auto-generated FAQs and powerful chatbots. With structured documentation, it also becomes possible to track real-world user behavior patterns anchored in specific pain points or product features — then close the loop by using that data to train AI tools based on users’ real-world needs and queries.

We’re already using Topic-based NLP to generate accurate documentation summaries, for instance, making it easier than ever for users to find the exact information they need. And because Conveyer’s contextual approach also enables images and video assets to be associated with a specific idea block, we’re able to deliver salient media-rich content in ways that today’s conventional LLMs and NLP tools can’t match. 

🔮The future of product documentation

Because Conveyer is powered by contextual algorithms acting on brands’ own product information, not generic NLP tools trained on large public datasets, we’re able to deliver robust model-specific results for today’s most demanding brands. 

That, in turn, enables brands to give customers near-instant access to dependable, manufacturer-approved information — and to use granular engagement data to drive innovation, improve the user journey, and unlock new revenue channels. 

Our proprietary AI solutions help brands to unlock the full potential of their product information — while ensuring they stay in full control of their customer relationships. What comes next? That would be telling, but we have big plans. ➡Get in touch to find out more about what lies in store.