[recognize entities] fix skipping entities beyond a certain index in …

…the query (#1351)

There was a bug in limiting the n-grams used to recognize entities where
beyond a certain number of words in the query, the rest of the query was
not being used in recognition

internal/server/recon/golden/recognize_entities/result.json

@@ -92,6 +92,53 @@
           "span": "jose are chirpy"
         }
       ]
+    },
+    "what genes are associated with the rs13317 , rs1826962 , rs790314 , rs2801952 , rs1814149": {
+      "items": [
+        {
+          "span": "what genes are associated with the"
+        },
+        {
+          "span": "rs13317",
+          "entities": [
+            {
+              "dcid": "bio/rs13317"
+            }
+          ]
+        },
+        {
+          "span": ", rs1826962 ,",
+          "entities": [
+            {
+              "dcid": "bio/rs1826962"
+            }
+          ]
+        },
+        {
+          "span": "rs790314",
+          "entities": [
+            {
+              "dcid": "bio/rs790314"
+            }
+          ]
+        },
+        {
+          "span": ", rs2801952 ,",
+          "entities": [
+            {
+              "dcid": "bio/rs2801952"
+            }
+          ]
+        },
+        {
+          "span": "rs1814149",
+          "entities": [
+            {
+              "dcid": "bio/rs1814149"
+            }
+          ]
+        }
+      ]
     }
   }
 }

internal/server/recon/golden/recognize_entities_test.go

@@ -45,6 +45,7 @@ func TestRecognizeEntities(t *testing.T) {
 					// should not recognize the first "me" but should recognize "gene me"
 					// and "mesh descriptor genes"
 					"tell me about the gene me and the MeSH descriptor genes",
+					"What genes are associated with the rs13317 , rs1826962 , rs790314 , rs2801952 , rs1814149",
 				},
 				"result.json",
 			},

internal/server/recon/recognize.go

@@ -257,9 +257,12 @@ func getId2Span(query string) map[string]map[string]struct{} {
 	spanTokens := strings.Split(query, " ")
 	for i := range spanTokens {
 		span := ""
-		maxNGramLength := int(math.Min(float64(len(spanTokens)), reconNGramLimit+1))
+		// This is the index in the list of tokens to end at when making n-grams,
+		// which should be either the end of the list of tokens or when the max n is
+		// reached for the n-grams, whichever comes first.
+		maxNGramIdx := int(math.Min(float64(len(spanTokens)), float64(reconNGramLimit+i)))
 		// make n-grams from the span tokens
-		for j := i; j < maxNGramLength; j++ {
+		for j := i; j < maxNGramIdx; j++ {
 			span = span + " " + spanTokens[j]
 			span = strings.TrimSpace(span)
 			id := getReconName(span)