A business’s most valuable assets are its conversations – both internal conversations amongst employees and external revenue-generating conversations with customers. Natural human conversations are messy, recall is different person to person, and companies suffer from inefficiency due to a lack of automation. Millions of business conversations take place every day that do not leverage the power of conversational intelligence technology to automate and optimize productivity-and-growth driving actions. This disjointed communication experience results in lost productivity and lost revenue.
Why not use a Conversational Intelligence API to solve the problem?
Symbl’s Conversational Intelligence API allows businesses to analyze business conversations and drive automated, real-time actions such as scheduling a follow-up, categorizing a conversation transcription, or presenting in-conversation recommendations. In other words, we provide companies the ability to amplify these interactions and build extraordinary customer experiences in any channel – be it voice, video, or text.
The Symbl API Advantage
With the Symbl API, we make it easy to integrate best-in-class Conversational Intelligence into your existing product. With Symbl you will be able to:
Step 1: Select the channels where conversations happen Step 2. Select the way you want to ingest the conversation stream from that channel Step 3: Extract your transcript and conversation insights
Symbl’s APIs make it easy for developers to focus on the experience of their customers by:
providing contextual intelligence beyond just speech to text and sentiment analysis
supplying a REST interface for Text and Voice Input and supports SIP, PSTN, Websocket, and audio files in real-time and async
requiring no training data or trigger words for domain specificity
enabling multiple use case – e.g., meetings, sales, support, legal, edtech, and healthcare
offering out-of-the-box integrations with work tools and extensible webhooks to automate workflows
Every business today produces a lot of conversational data in many forms like voice, video, recordings, live meetings, sales, and customer service interactions. All this valuable data is sometimes just stored in data centers, costing customers a lot of money and leading to no great value. This data could be used more efficiently when it’s happening in real time rather than living in centers collecting digital dust. It’s as simple as building applets on IFTTT or zaps on Zapier.
Think of it like ‘IF — (specific) Conversation, THEN — (specific) Action.’
We’re building Symbl’s conversational intelligence platform to specifically enable simple, low-code workflows to derive outcomes from conversations immediately. Let’s explore a few areas where conversations can lead to instant actions and help automate processes .
Tools like Slack, Zoom, Loom, Jira, and Google Drive have made business conversations and interactions easier, whether employees are working together in the same office space or remotely.
Here are a few workflow examples to boost workplace productivity through real-time actions in natural conversations:
Automatically assign a task to someone on Slack/Jira from a meeting or stand-up
Talk about sharing a file in a meeting and have it sent automatically to Google Drive
Make all your team demos searchable with transcripts and topics
Instantly share updates from interviews and stand-ups
Many sales conversations happen on conferencing platforms or phones that require tracking on CRM systems and immediate follow-ups.
Here are a few easy workflow examples that streamline sales pipelines in real-time without manually entering data:
Schedule any follow-up calls automatically on the calendar
Update opportunity-specific details to the CRM instantly
An aggregated view of real-time sales discussions on specific topics, mention of competition, etc.
Automatically progress leads to the life-cycle stage to set up demos, reviews, etc.
At high volumes, it’s hard to maintain the best customer experience and agent training quality in contact centers. But with continuous conversational intelligence, contact centers can reduce average handling time and increase customer experience.
Add these additional workflows to provide an exceptional experience for customers and contact center employees:
Use conversations to automatically enable document verification with RPA
Set up and communicate customer visits and appointments automatically
Build a live, searchable knowledgebase using historic customer service and on-going support conversations
Automate backend processes with no/minimal user intervention such as invoice generation, collection, etc with RPA
Automatically recommend relevant knowledge base articles to help agents have a smarter conversation
There are endless possibilities across industries and use cases of using workflows that can be triggered from conversations. Early customers and developers on our API platform are currently building similar workflows. We want to enable our customers and developers to pre-build, low-code steps for similar workflows. We’ll soon publish some of them as steps for Flows on the Flow Manager (beta).
Here are some sample Flow implementations we’re excited about:
Customer service conversation using Twilio Voice → Extract insights → Update on Salesforce CRM
Tech teams on any meeting platform→ Extract Insights & Actions → Update Jira tickets
As we think about how to solve the problem of natural human conversation understanding and how to extract information from unstructured conversations, the approach to solving this problem can make a significant impact on when and how we will end up getting there.
Before jumping into the application of Deep Learning and NLP solutions for tackling the above problem, let’s establish the inherent problem present in the conversation intelligence.
Compared to news articles and other kinds of free-flowing text available on the internet, conversations have a variety of different issues to be solved from an engineering, mathematical and scientific standpoint.
Conversations are complex, and so is the data. Why?
There are a variety of reasons why the data derived from conversations are difficult to collect and use to extract information. Here are a few examples of why that could be:
Conversations are random and dependent on state and user context
Meaning in conversations can be hierarchical, can be communicated in bits and pieces at irregular intervals with no inference
Conversations consist of topics that break at uneven points and can be disentangled. For example, we may speak about my trip to India for a while and then switch back to another experience. This raises many problems in understanding co-referencing and other aspects of structuring them.
Conversational data is heavily subjective to the user present in the meeting. For a third party witnessing the conversation, it may take a long time and a lot of effort to comprehend.
Error propagation is possible due to heavy bias and dependency on speech recognition
Context is an integral part of the conversation and is highly uncertain. It is synonymous with uncertainty similar to a non-zero sum game where it can’t be modeled as a fixed vector at any point in time. The modeling is always dependent on uncertain conditions.
The inference AI model has to learn from a small amount of initialization data
Conversation intelligence stands apart from traditional NLP where the hypothesis, situation and inference communicated by the user are clearly stated and explained in a given flow. Conversational data is all over the place and has a clear mismatch with the existing embeddings and pre-trained models used on the scale by the NLP community. This makes conversation intelligence beyond a curve-fitting problem and it can be transposed as a problem of making decisions under uncertainty and incomplete information.
Deep learning problems when trained with advanced mechanisms like attention and multi-head attention tend to perform well on certain NLP tasks. But they don’t really work for conversations as they fall prey to the problems of the train-test mismatch of conversation vs articles/documents. Second, it’s computation-intensive and costly to inject context.
Context can be intuitively understood as “What led to this point in the conversation and where may it lead next?” As we look for answers, let’s dig a little deeper into what both sides of the spectrum can fetch us, and a few of the approaches and aspects that hybrid learning might prove worthy.
Need additional help? You can refer to our API Docs for more information and view our sample projects on Github.
Business Intelligence (B.I.) is the infrastructure that a business uses to collect, organize and manage all its data. Everything from simple spreadsheets to the more thorough dashboards, all fall under the umbrella of B.I. While B.I. has been a fundamental part of operations and business strategy decisions, it is traditionally used to present data in a more readable way. As Michael F. Gorman, professor of operations management and decision science at the University of Dayton in Ohio, said in an article published by CIO Magazine, “[Business Intelligence] doesn’t tell you what to do; it tells you what was and what is.”
As this space has evolved with the adoption of Artificial Intelligence(AI), we’ve identified some of the ways in which AI enhances the simple business intelligence tools into something more powerful.
Where is the B.I train headed?
Among several important developments thanks to the adoption of machine intelligence, one important aspect is the expanding sources of data. While businesses take into consideration most of the new technologies like IoT, click tracking, and Robotic process automation (RPA) systems that provide waterfalls of useful intel, a lot of effort is spent to discover additional data sources.
With the increase in the number of data sources, the next stop on the data train is at the junction of big data technologies. There is a need for advanced AI to analyze large amounts of data and there are efforts being taken to address it. Apache Hadoop and Spark are some of the most popular open-source frameworks to store and process big data in a distributed environment.
One of the most commonly adopted new business intelligence approaches due to significant ROI potential is Predictive business analysis. This is where large amounts of historical data is analyzed to predict future outcomes – one of the most common use cases is the implementation of Next Best Action in call centers. For example, a call center agent might use the historical data from past appointments to reach out to the customer to book their next appointment.
As we continue to explore the avenues of real-time B.I, conversation analysis systems will become crucial so that these predictions based on historical data can complement with real-time customer conversation experiences
Conversational Intelligence – the third eye of B.I.
Sophisticated conversational analysis platforms can contribute immensely to the usefulness of the data sources. Speech to Text platforms have been in existence for a while and are in fact starting to become commoditized, but extracting the nuggets of information and insights can change how the real-time B.I tools deliver value to organizations. This capability extends B.I beyond “what is” and “what was” and is more indicative of “what can be done”. This is where B.I becomes action-oriented.
Prophecies for the unleashed
The influx of data seen by businesses encourages the use of proactive, real-time systems for reporting and analysis that can help with alerts and updates. Getting the meta-conversation data source: Conversations are the last mile of data getting lost today and is heavily underutilized. The meta-level information on customer conversation statistics can give horsepower to the B.I systems for real-time, actionable intelligence – for example understanding customer sentiments can explain the outcomes of the exchange.. Imagine having this flow into the customer support organization to influence real-time call center conversations resulting in service recovery. For supervised learning approaches, it is important to highlight the dependency of the quality of an AI model on the quality of the data, which is why the exploitation of all the available data should be utilized. Sources such as meetings, client interactions website visitors, chatbots, feedback and review forums, and demand trends are a few gold mines of data.
Consider the infamous 80-20 rule: It is an axiom of business management that “80% of sales come from 20% of clients”. With so much data, it is important to identify what is most important. Whether this identification is done through AI or domain experts within the company it is definitely something to keep in mind.
Beyond visual representation and the Role of Natural language generation (NLG) in B.I: New developing branches of AI like NLG have a major impact on the usefulness and accessibility. Jon Arnold, Principal of J Arnold and Associates, in his article for Enterprise Management 360 says- “A key reason why BI is a strong use case is that these platforms provide visual representations of the data, but this isn’t always helpful for workers. Not all forms of data can be easily visualized, and visual outputs aren’t always enough. Sometimes a written analysis is what’s needed, and other times voice is the format that works best.”
Stay tuned for a blog that goes beyond traditional B.I and how cognitive RPA and B.I are changing the way enterprises work. If you are interested in learning how Symbl helps B.I. teams collect conversational data by setting up the metadata and determining the right design patterns, contact us for a demo.
Suppose I give you a simple problem statement: Predict the next number in the sequence
2, 4, 6, 8, 10, __?
Easy peasy right? I’m not going to insult your intelligence by telling you the answer. I mean, come on.
Now let’s take another sequence :
8, 13, 21, 34, 55, __?
Not so easy as the first one, is it? Well, if you’re still scratching your head, it’s a Fibonacci sequence, where the next number is the sum of the previous two numbers. And the next number is 89.
Now how about this sequence :
0, 2, 4, __?
You might be a bit confused now. Is the next number 6, where the sequence is just adding 2 to the previous number? Or is it 8, where the sequence is just powers of 2? Can you decide?
You can’t really, can you? But why can’t you make a decision? Suppose I add another number to the sequence :
0, 2, 4, 8, __?
Okay so now we know. It’s going to be 16, as we found out the pattern in the sequence was exponents of 2.
Why couldn’t we decide earlier? And why did adding one more instance instantly solve our dilemma? Well, to put it plainly, you got more data to work with, and then could extract a pattern.
This is precisely what Deep Learning models do. You feed it data and beg the wizards hiding inside the models to figure out patterns that make everything sensible. (Deep Learning is fun!)
Congratulations, you’re now a Data Scientist. Start brushing up that resume.
Why Go for Deep Learning?
One word. Accuracy.
When the problem statement is complex, there will be more exceptions than rules. This rules out a simple rule-based/algorithm-based system. And even when resorting to Machine Learning (ML), the patterns a Data Scientist would find, that would generate the feature set required for these models, would have exceptions that can be very hard to tackle.
Entering the world of Deep Learning and Neural Networks.
As mentioned earlier, Deep Learning (DL) models work on data (labeled data). And the model architecture you’ve designed, learns patterns from the data, all by itself. And it does this by constantly iterating over the training data, making predictions at each iteration, and finding out what the error between the prediction and the ground truth is. This error signal is then used to update the parts in the network that ‘learn’ ( which we call ‘weights’ of the neurons). This process is repeated until the error signal between the predicted truth and the ground truth is minimal.
Now you must be thinking where are the wizards that I mentioned earlier. Not to burst your bubble, but there are no wizards inside the neural net, it’s just math, particularly calculus. Sorry.
Neural nets and conversations – the ensemble way!
Now coming to conversations. conversational data are not something that you can find in loads, on the internet. Well, you can, but most of them would be either simulated, in the form of chat corpora (which is way different from how people speak), or they wouldn’t fit the use case that we’re trying to solve. And for a neural net, it always needs data.
We can’t just go back to standard ML and rule-based approaches. They would be as helpful as having a toothbrush when you’re tasked with cleaning an entire floor. I mean, you can do it, but at what cost?
And I don’t take hallucinogenic drugs, so I’m under no delusions that a single DL model trained on limited data can give us godly results.
This is where “ensembling” various DL models and other methods can be used for exponential payoffs. Each neural net would learn different things from the data it’s exposed to. Then it’s just a matter of combining these multiple “weak learners” (“weak” because they each learn specific things, but not the entire thing) to get the best possible model.
Now, even this mighty ensemble might fail to spot simple patterns, and so what do we do for that? Well, now since our floor is 95% clean thanks to the deep learning ensemble, we can finally make use of our toothbrush and apply the standard age-old techniques to weed out the rest of the dirt.