Update README.md

README.md

@@ -187,14 +187,23 @@ Validation values obtained from tensorboard from epoch 2399*:
 
 ## Analyzing Multi-Accented Inferences
 
-Here we show aligned spectrograms with phonemes from the same inference generated with the four accents:
+Here we show aligned spectrograms with phonemes from the same inference generated with the four accents. The following spectrograms correspond to the sentence *La seva gerra sembla molt antiga*.
+This utterances shows three phonetic particularities that differentiate the Catalan variants.
 
+In the Balearic and Valencian accents, [t] is pronounced, whereas it is not pronounced in the other two accents. 
+It can be seen that the **-v-** in *seva* is pronounced as [] in Central accent while as [w] in Valencian.
+Also, it is observed that the // in *gerra* has an affricate pronunciation ([]) in Valencian.
+
+*Balearic inference*
 ![Balearic Inference](./inference_specs/infer_ba.png)
 
+*Central inference*
 ![Central Inference](./inference_specs/infer_ce.png)
 
+*North-Western inference*
 ![North-Western Inference](./inference_specs/infer_nw.png)
 
+*Valencian inference*
 ![Valencian Inference](./inference_specs/infer_va.png)