seaborn heatmap - x axis - repeated values

Question

I'm in trouble creating a heatmap using a CSV file.

csv data is in a format like below

here is a code

years = np.array(datadf.PublicationYear)
sns.set(font_scale=2)
wordlist = ['greenhouse_gas', 'pollution', 'resilience', 'urban','city', 'environmental_impacts', 'climate_change', 
            'adaptation','mitigation','carbon', 'ghg_emissions','sustainable','sustainability','lca']
word_tuples = [('urban','city','urban'), ('greenhouse_gas','ghg_emissions','greenhouse_gas'),('sustainable','sustainability','sustainability')]
use_wordlist = True
word_number = 30
freqdata = []
agg_keys = []
for i in np.arange(len(abstrct)):
    ngram_model = Word2Vec(ngram[[abstrct[i]]], size=100, min_count=1)
    ngram_model_counter = Counter()
    for key in ngram_model.wv.vocab.keys():
    if key not in stoplist:
        if use_wordlist:
            if key in wordlist:                   
                if len(key.split("_")) > N:
                    ngram_model_counter[key] += ngram_model.wv.vocab[key].count
        else:
            if len(key.split("_")) > N:
                    ngram_model_counter[key] += ngram_model.wv.vocab[key].count


freqdf = pd.DataFrame(ngram_model_counter.most_common(word_number))
if len(freqdf.index) == 0:
    freqdf[0] = wordlist
    freqdf[1] = 0


for w in word_tuples:
    if w[0] in wordlist and w[1] in wordlist:
        f = 0
        drops_w = []

        for j in np.arange(len(freqdf.index)):
            if freqdf.iloc[j][0] == w[0] or freqdf.iloc[j][0] == w[1]:

                f += freqdf.iloc[j][1]
                drops_w.append(j)

        freqdf = freqdf.drop(drops_w, axis = 0)        
        append_data = pd.DataFrame({0:[w[2]],1:[f]})
        freqdf = freqdf.append(append_data,ignore_index=True)
        freqdf = freqdf.reset_index(drop=True)


#Normalizing the frequency by the total number of non-stopword tokens 
freqdf['prob'] = freqdf[1]/(len(abstrct[i]))

agg_keys += np.array(freqdf[0]).tolist()
freqdata.append(freqdf)  



unqkeys = np.unique(np.array(agg_keys))
matrix = np.zeros([unqkeys.size,len(freqdata)])
for i in np.arange(years.size):
for j in np.arange(unqkeys.size):

    for k in np.arange(len(freqdata[i].index)):

        if freqdata[i].iloc[k][0] == unqkeys[j]:
            matrix[j,i] = freqdata[i].iloc[k]['prob']


fig, ax = plt.subplots(figsize = (40, 26))
ax = sns.heatmap(matrix, annot = False,linewidths = .9,cmap = 'Blues' ,cbar_kws={'label': 'Scaled Frequency'})
ax.figure.axes[-1].yaxis.label.set_size(35)
ax.grid(False)
ax.set_yticks(np.arange(len(unqkeys))+0.5) #Adding 0.5 offset
ax.set_xticks(np.arange(len(years))+0.5)
ax.set_yticklabels(unqkeys,rotation= 0, fontsize = 34.0)
ax.set_xticklabels(years,rotation='vertical', fontsize = 35.0)
ax.set_xlabel('Publication Year', fontsize = 40.0, labelpad = 40)
ax.set_ylabel('Keywords',fontsize = 40.0, labelpad = 5)
plt.show()
image_path =  os.path.join(base_dir, 'plots/') + data_dir + '-heatmap.jpg'
fig.savefig(image_path)

It generates a graph below. How could it make x axis like [2016,2017,2018,2019,2020,2021]