你不会还天真的以为只有人类会作诗吧！

下一个诗仙、诗圣或许根本不是人！...

《点绛唇》

人静风清，兰心蕙性盼如许。

夜寒疏雨，临水闻娇语。

佳人多情，千里独回首。

别离后，泪痕衣袖，惜梦回依旧。

这是一首电脑写的诗

A wounded deer leaps highest.

I’ve heard the daffodil.

I’ve heard the flag to-day.

I’ve heard the hunter tell；

'Tis but the ecstasy to death，

And then the brake is almost done.

And sunrise grows so near,

Sunrise grows so near,

That we can touch the despire and frenzied hope of all the ages.

这也是一首电脑写的诗

《相思》

Missing You

红豆生南国，（＊ZPPZ)

Red berries born in the warm southland.

春来发几枝？(PPZZP)

How many branches flush in the spring?

愿君多采撷，(*PPZZ)

Take home an armful, for my sake,

此物最相思。(*ZZPP)

As a symbol of our love.

这是一首电脑翻译出的诗

连平仄音都分得出

就算没游山玩水，

它也能写出：

明月松间照，清泉石上流。

就算没喝过醉过，

它也能写出：

常记溪亭日暮，沉醉不知归路。

就算没谈过异地恋，

它也能写出：

一种相思，两地闲愁；

就算没经历改朝换代，

它也能写出：

国破山河在，城春草木深。

语言不过是一种工具，大多数人都用不好的工具，千丝万缕的小情绪经常找不到合适的片语只言来形容。

诗是语言界的名门。规矩多多，律诗的格律貌似最严格，篇有定句（每首八句），句有定字（五字或七字），字有定声（平仄相对），联有定对（中间两联对仗），绝句平仄对仗没有律诗那么绝，每首四句，有五言，七言两种，简称五绝，七绝。

情绪要在这个规则下溢出，不少人为了做几首诗咬文嚼字，为难自己，就像本来好好的脚被裹成三寸金莲。

这么说来，AI貌似比人类更适合作诗，随机应变不逾矩又能拨动心弦。

别说它没感情，也许它信手拈来就能感动天下人。反正这个世界向来愧对天才。RNNGG是这么作诗的：

Figure 1:

Poem generation with keywords spring，lute，and drunk.The keywords are expanded into phrases using a poetic taxonomy. Phrases are then used to generate the first line. Following lines are generated by taking into account the representations of all previously generated lines.

Figure 2:

Generation of the (j+1)th character wj+1 in the (i + 1)th line Si+1. The recurrent context model (RCM) takes i lines as input (represented by vectors v1 , . . . , vi ) and creates context vectors for the recurrent generation model (RGM). The RGM estimates the probability P(wj+1|w1:j,S1:i).

RCM(Recurrent Context Model)

RGM(Recurrent Generation Model)

Figure 3:

Convolutional sentence model for 7-char quatrain. The first layer has seven vectors,one for each c-hacter. Two neighboring vectors are merged to one vector in the second layer with weight matrixC’2. . In other layers, either two or three neighboring vectors are merged.