Note: The Virtual Agent can be downloaded here » Virtual Agent SDK (zip) (VirtualAgentSDK\Samples\Alchemy)
The Alchemy sample from the Virtual Agent SDK includes two Virtual Agent classes. The Pre-Conversation Virtual Agent gets the top 5 keywords from the guest’s question and adds this information to the session’s metadata in Chime. The Post-Conversation Virtual Agent does a sentiment analysis on the chat messages from the Chime session. The sentiment type (positive or negative) and score (how strongly the sentiment was) are added to the Chime session metadata.
The following examples use these references. These examples also use a reference to the Alchemy sdk.dll.
Alchemy® Integration (3)
using System;
using System.Collections.Generic;
using System.IO;
using System.Linq;
using System.Text;
using System.Xml.Linq;
using System.Xml.Serialization;
The following code shows you how to authenticate with the Alchemy AI service. The only credential this requires is your Alchemy API key which can be requested from http://www.alchemyapi.com/api/register.html.
AlchemyAPI alchemyAPI = new AlchemyAPI.AlchemyAPI();
public void ConnectToAlchemy()
{
try
{
alchemyAPI.SetAPIKey(apiKey);
isConnectedToAlchemy = true;
}
catch (Exception ex)
{
Console.WriteLine(ex.ToString());
}
}
One of the things Alchemy API is very good at is analyzing a body of text. This first example will show you how to leverage this capability to extract topic keywords from some text, ranked by their relevance as perceived by the Alchemy service. Note, Alchemy responses are in XML format. For brevity, the RankedKeywords class that is used to bind to the deserialized response is not included, but you can construct your own class from the resources provided by Alchemy on this page (scroll to the bottom Response Format (XML) section) http://www.alchemyapi.com/api/keyword/textc.html. I’ve also excluded the TopFiveKeywords method, this simply string-ifies the text field of the first five keywords.
public RankedKeywords.Response GetRankedKeywords(String text)
{
RankedKeywords.Response status = new RankedKeywords.Response();
try
{
if (alchemyAPI != null)
{
RankedKeywords.results response = new RankedKeywords.results();
XmlSerializer serializer = new XmlSerializer(response.GetType());
var xml = alchemyAPI.TextGetRankedKeywords(text);
Console.WriteLine(xml);
object deserialized = serializer.Deserialize(ToStream(alchemyAPI.TextGetRankedKeywords(conversationText)));
response = (RankedKeywords.results)deserialized;
status.Result = response;
status.Status = "SUCCESS";
Console.WriteLine("Ranked Keyword Count: " + status.Result.keywords.Count());
Console.WriteLine("Top 5 Keywords: " + TopFiveKeyWords(status.Result.keywords));
}
}
catch (Exception ex)
{
status.Result = null;
status.Status = "ERROR: " + ex.ToString();
Console.WriteLine(ex.ToString());
}
return status;
}
This second example will show you how to use Alchemy API to retrieve a sentiment analysis of a body of text. Alchemy uses the following data points to report sentiment analysis:
Once again, for brevity we have excluded the class source code that binds to the Alchemy XML response. This documentation shows you how the response will be formed http://www.alchemyapi.com/api/sentiment/textc.html.
public SentimentAnalysis.Response GetSentiment(string conversationText)
{
SentimentAnalysis.Response status = new SentimentAnalysis.Response();
try
{
if (alchemyAPI != null)
{
SentimentAnalysis.results response = new SentimentAnalysis.results();
XmlSerializer serializer = new XmlSerializer(response.GetType());
object deserialized = serializer.Deserialize(ToStream(alchemyAPI.TextGetTextSentiment(conversationText)));
response = (SentimentAnalysis.results)deserialized;
status.Result = response;
status.Status = "SUCCESS";
if (status.Result != null && status.Result.docSentiment != null)
{
if (status.Result.docSentiment.type != null)
{
Console.WriteLine("Sentiment type: " + status.Result.docSentiment.type);
}
Console.WriteLine("Sentiment score: " + status.Result.docSentiment.score);
var mixed = status.Result.docSentiment.mixed;
Console.WriteLine("Mixed sentiment: " + (mixed == 1 ? "True" : "False"));
}
}
}
catch (Exception ex)
{
Console.WriteLine(ex.ToString());
}
return status;
}
Have a question about Chime integration with Alchemy®? Ask one of our developers at InstantDev@instant-tech.com.
