/// <summary>
/// Return a document for a query with the given id.
/// </summary>
static Document getQueryById(String sid, DocumentCollection queryCollection)
{
int queryId = int.Parse(sid);
var doc = queryCollection.documentById(queryId);
if (doc == null)
{
Console.WriteLine("Error: Query with ID {0} doesn't exist!", queryId);
return null;
}
return doc;
}
/// <summary>
/// Update the IR system by building a search index using the
/// given document collection.
/// </summary>
public void update(DocumentCollection documents)
{
this.documents = documents;
foreach (var doc in documents.getDocuments())
{
// Create a local (document specific) term collection.
Dictionary<String, int> documentTerms = new Dictionary<String, int>();
foreach (var rawWord in doc.getTerms())
{
var word = stemmer.stem(rawWord);
if (stopWords.isStopWord(word)) { continue; }
int frequency;
if (documentTerms.TryGetValue(word, out frequency))
{
documentTerms[word] = frequency + 1;
}
else
{
documentTerms[word] = 1;
}
}
// Merge the local term collection with the global term document matrix.
foreach (var termOccurence in documentTerms) {
var term = terms.getTerm(termOccurence.Key);
term.addOccurenceForDocument(termOccurence.Value, doc.id);
}
}
}
static void Main(string[] args)
{
Console.WriteLine("Welcome to ir-project command line terminal interface.");
Console.WriteLine("----");
String input;
var loadDocuments = new Command("load documents", "filepath");
var loadQueries = new Command("load queries", "filepath");
var loadRelevance = new Command("load relevance", "filepath");
var runNewQuery = new Command("run text query", "text");
var runQuery = new Command("run query", "query id");
var showTerm = new Command("show term", "term");
var showDocument = new Command("show document", "document id");
var showQuery = new Command("show query", "query id");
var setScheme = new Command("set scheme", "bm25|pivoted");
Command[] commands = { loadDocuments, loadQueries, loadRelevance, runNewQuery, runQuery, showTerm, showDocument, showQuery, setScheme };
var documentCollection = new DocumentCollection();
var queryCollection = new DocumentCollection();
List<RelevanceJudgement> relevance = null;
var engine = new InformationRetriever();
// Weighting schemes.
var bm25Scheme = new BM25Scheme(/*k1=*/2.0f, /*b=*/0.75f);
var pivotedScheme = new PivotedNormalisationScheme(/*s=*/0.5f);
WeightingScheme scheme = bm25Scheme;
do
{
Console.Write("> ");
input = Console.ReadLine();
if (input == null) { input = ""; }
if (input == "exit" || input == "quit") { break; }
if (input == "help") {
Console.WriteLine("Terminal UI for the ir-project IR system.");
Console.WriteLine("Available commands:");
Console.WriteLine(" exit");
Console.WriteLine(" quit");
Console.WriteLine(" help");
foreach (var command in commands)
{
Console.WriteLine(" {0}: <{1}>", command.command, command.argumentName);
}
continue;
}
String matchedCommand = "";
String matchedArgument = null;
foreach (var command in commands) {
var match = command.matches(input);
if (match.Item1) {
matchedCommand = command.command;
matchedArgument = match.Item2.Trim();
break;
}
}
switch (matchedCommand) {
case "load documents":
{
var data = readFile(matchedArgument);
if (data != null)
{
Console.WriteLine("Loading documents... ");
var documents = DataImporter.parseDocuments(data);
Console.WriteLine("Loaded {0} documents!", documents.Count);
Console.Write("Constructing the term-document indexing datastructures... ");
Stopwatch sw = new Stopwatch();
sw.Start();
documentCollection = new DocumentCollection(documents);
engine.update(documentCollection);
sw.Stop();
Console.WriteLine("Done ({0} ms)!", sw.ElapsedMilliseconds);
}
break;
}
case "load queries":
{
var data = readFile(matchedArgument);
if (data != null)
{
Console.WriteLine("Loading queries");
var queries = DataImporter.parseDocuments(data);
queryCollection = new DocumentCollection(queries);
Console.WriteLine("Loaded {0} queries!", queries.Count);
}
break;
}
case "load relevance":
{
var data = readFile(matchedArgument);
if (data != null)
{
Console.Write("Loading relevance judgements... ");
relevance = DataImporter.parseRelevance(data);
Console.WriteLine("Loaded {0} relevance judgements!", relevance.Count);
}
break;
}
case "run text query":
{
var query = engine.createQuery(new Document(matchedArgument, 0));
var results = engine.executeQuery(query, scheme);
Console.WriteLine("Found {0} results:", results.Count);
foreach (var result in results)
{
Console.WriteLine(" document id: {0}, similarity score: {1}", result.documentId, result.similarity);
}
break;
}
case "run query":
{
var doc = getQueryById(matchedArgument, queryCollection);
if (doc == null)
{
break;
}
Console.WriteLine("Query with id {0}, text: '{1}'", doc.id, doc.value);
Stopwatch sw = new Stopwatch();
sw.Start();
var query = engine.createQuery(doc);
var results = engine.executeQuery(query, scheme);
sw.Stop();
Console.WriteLine("Found {0} results in {1} ms:", results.Count, sw.ElapsedMilliseconds);
foreach (var result in results)
{
Console.WriteLine(" document id: {0}, similarity score: {1}", result.documentId, result.similarity);
}
// Compute precision/recall if possible.
if (relevance != null) {
var relevantDocuments = relevance.Where(x => x.queryId == doc.id).Select(x => x.documentId).ToArray();
var relevantSet = new HashSet<int>(relevantDocuments);
List<double> precision = new List<double>();
List<double> recall = new List<double>();
//add initial precision and recall values
precision.Add(100);
recall.Add(0);
double noOfRelevantDocs = relevantDocuments.Count();
double relevantDocCount = 1;
double resultCount = 1;
foreach (var result in results)
{
if (relevantSet.Contains(result.documentId))
{
recall.Add((relevantDocCount / noOfRelevantDocs)*100);
precision.Add((relevantDocCount / resultCount)*100);
relevantDocCount++;
}
resultCount++;
}
Console.WriteLine("Precision | Recall (graph is shown in the browser):");
for(int i = 0; i < precision.Count; i++)
{
Console.WriteLine(" {0} | {1}", precision[i], recall[i]);
}
PrecisionRecallGraph.show(doc.id, precision, recall);
}
break;
}
case "show term":
{
var term = engine.terms.findTerm(matchedArgument);
if (term == null)
{
Console.WriteLine("Error: Invalid term '{0}'", matchedArgument);
break;
}
Console.WriteLine("Term '{0}', global frequency: {1}, occurences:", matchedArgument, term.getGlobalFrequency());
foreach (var occurence in term.getOccurences())
{
Console.WriteLine(" document id: {0}, frequency: {1}", occurence.documentId, occurence.frequency);
}
break;
}
case "show document":
{
int docId = int.Parse(matchedArgument);
var doc = documentCollection.documentById(docId);
if (doc == null)
{
Console.WriteLine("Error: Document with ID {0} doesn't exist!", docId);
break;
}
Console.WriteLine("Document with id '{0}', terms:", doc.id);
foreach (var term in engine.terms.termList)
{
var occurences = term.getOccurences().Where(x => x.documentId == doc.id);
// Usually this just one occurence.
foreach (var occurence in occurences)
{
Console.WriteLine(" term '{0}', frequency: {1}", term.value, occurence.frequency);
}
}
break;
}
case "show query":
{
var doc = getQueryById(matchedArgument, queryCollection);
if (doc == null)
{
break;
}
Console.WriteLine("Query with id {0}, text: '{1}', terms:", doc.id, doc.value);
var query = engine.createQuery(doc);
foreach (var term in query.terms)
{
Console.WriteLine(" term '{0}', frequency: {1}", term.term.value, term.frequency);
}
break;
}
case "set scheme":
{
if (matchedArgument == "bm25")
{
scheme = bm25Scheme;
Console.WriteLine("Using BM25.");
}
else if (matchedArgument == "pivoted")
{
scheme = pivotedScheme;
Console.WriteLine("Using Pivoted Normalisation.");
}
else
{
Console.WriteLine(" unknown weighting scheme '{0}'", matchedArgument);
}
break;
}
default:
Console.WriteLine("Error: Unknown command '{0}'!", input);
break;
}
} while (input != "exit" && input != "quit");
Console.WriteLine("---");
}